{"version": 1, "context": "ComputeGraph { inputs: [GraphInput { name: \"vit.image.patches\", kind: Host, value: ValueId(0), shape: TensorShape([2, 729, 588]) }, GraphInput { name: \"vit.embedding.weight\", kind: Parameter, value: ValueId(1), shape: TensorShape([588, 1152]) }, GraphInput { name: \"vit.embedding.bias\", kind: Parameter, value: ValueId(2), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.position\", kind: Parameter, value: ValueId(5), shape: TensorShape([1, 729, 1152]) }, GraphInput { name: \"vit.encoder.layer0.attention_norm.scale\", kind: Parameter, value: ValueId(7), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer1.attention_norm.scale\", kind: Parameter, value: ValueId(8), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer2.attention_norm.scale\", kind: Parameter, value: ValueId(9), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer3.attention_norm.scale\", kind: Parameter, value: ValueId(10), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer4.attention_norm.scale\", kind: Parameter, value: ValueId(11), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer5.attention_norm.scale\", kind: Parameter, value: ValueId(12), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer6.attention_norm.scale\", kind: Parameter, value: ValueId(13), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer7.attention_norm.scale\", kind: Parameter, value: ValueId(14), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer8.attention_norm.scale\", kind: Parameter, value: ValueId(15), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer9.attention_norm.scale\", kind: Parameter, value: ValueId(16), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer10.attention_norm.scale\", kind: Parameter, value: ValueId(17), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer11.attention_norm.scale\", kind: Parameter, value: ValueId(18), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer12.attention_norm.scale\", kind: Parameter, value: ValueId(19), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer13.attention_norm.scale\", kind: Parameter, value: ValueId(20), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer14.attention_norm.scale\", kind: Parameter, value: ValueId(21), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer15.attention_norm.scale\", kind: Parameter, value: ValueId(22), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer16.attention_norm.scale\", kind: Parameter, value: ValueId(23), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer17.attention_norm.scale\", kind: Parameter, value: ValueId(24), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer18.attention_norm.scale\", kind: Parameter, value: ValueId(25), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer19.attention_norm.scale\", kind: Parameter, value: ValueId(26), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer20.attention_norm.scale\", kind: Parameter, value: ValueId(27), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer21.attention_norm.scale\", kind: Parameter, value: ValueId(28), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer22.attention_norm.scale\", kind: Parameter, value: ValueId(29), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer23.attention_norm.scale\", kind: Parameter, value: ValueId(30), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer24.attention_norm.scale\", kind: Parameter, value: ValueId(31), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer25.attention_norm.scale\", kind: Parameter, value: ValueId(32), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer26.attention_norm.scale\", kind: Parameter, value: ValueId(33), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer0.attention_norm.bias\", kind: Parameter, value: ValueId(34), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer1.attention_norm.bias\", kind: Parameter, value: ValueId(35), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer2.attention_norm.bias\", kind: Parameter, value: ValueId(36), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer3.attention_norm.bias\", kind: Parameter, value: ValueId(37), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer4.attention_norm.bias\", kind: Parameter, value: ValueId(38), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer5.attention_norm.bias\", kind: Parameter, value: ValueId(39), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer6.attention_norm.bias\", kind: Parameter, value: ValueId(40), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer7.attention_norm.bias\", kind: Parameter, value: ValueId(41), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer8.attention_norm.bias\", kind: Parameter, value: ValueId(42), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer9.attention_norm.bias\", kind: Parameter, value: ValueId(43), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer10.attention_norm.bias\", kind: Parameter, value: ValueId(44), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer11.attention_norm.bias\", kind: Parameter, value: ValueId(45), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer12.attention_norm.bias\", kind: Parameter, value: ValueId(46), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer13.attention_norm.bias\", kind: Parameter, value: ValueId(47), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer14.attention_norm.bias\", kind: Parameter, value: ValueId(48), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer15.attention_norm.bias\", kind: Parameter, value: ValueId(49), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer16.attention_norm.bias\", kind: Parameter, value: ValueId(50), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer17.attention_norm.bias\", kind: Parameter, value: ValueId(51), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer18.attention_norm.bias\", kind: Parameter, value: ValueId(52), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer19.attention_norm.bias\", kind: Parameter, value: ValueId(53), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer20.attention_norm.bias\", kind: Parameter, value: ValueId(54), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer21.attention_norm.bias\", kind: Parameter, value: ValueId(55), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer22.attention_norm.bias\", kind: Parameter, value: ValueId(56), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer23.attention_norm.bias\", kind: Parameter, value: ValueId(57), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer24.attention_norm.bias\", kind: Parameter, value: ValueId(58), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer25.attention_norm.bias\", kind: Parameter, value: ValueId(59), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer26.attention_norm.bias\", kind: Parameter, value: ValueId(60), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer0.attention.qkv.weight\", kind: Parameter, value: ValueId(61), shape: TensorShape([1152, 3456]) }, GraphInput { name: \"vit.encoder.layer1.attention.qkv.weight\", kind: Parameter, value: ValueId(62), shape: TensorShape([1152, 3456]) }, GraphInput { name: \"vit.encoder.layer2.attention.qkv.weight\", kind: Parameter, value: ValueId(63), shape: TensorShape([1152, 3456]) }, GraphInput { name: \"vit.encoder.layer3.attention.qkv.weight\", kind: Parameter, value: ValueId(64), shape: TensorShape([1152, 3456]) }, GraphInput { name: \"vit.encoder.layer4.attention.qkv.weight\", kind: Parameter, value: ValueId(65), shape: TensorShape([1152, 3456]) }, GraphInput { name: \"vit.encoder.layer5.attention.qkv.weight\", kind: Parameter, value: ValueId(66), shape: TensorShape([1152, 3456]) }, GraphInput { name: \"vit.encoder.layer6.attention.qkv.weight\", kind: Parameter, value: ValueId(67), shape: TensorShape([1152, 3456]) }, GraphInput { name: \"vit.encoder.layer7.attention.qkv.weight\", kind: Parameter, value: ValueId(68), shape: TensorShape([1152, 3456]) }, GraphInput { name: \"vit.encoder.layer8.attention.qkv.weight\", kind: Parameter, value: ValueId(69), shape: TensorShape([1152, 3456]) }, GraphInput { name: \"vit.encoder.layer9.attention.qkv.weight\", kind: Parameter, value: ValueId(70), shape: TensorShape([1152, 3456]) }, GraphInput { name: \"vit.encoder.layer10.attention.qkv.weight\", kind: Parameter, value: ValueId(71), shape: TensorShape([1152, 3456]) }, GraphInput { name: \"vit.encoder.layer11.attention.qkv.weight\", kind: Parameter, value: ValueId(72), shape: TensorShape([1152, 3456]) }, GraphInput { name: \"vit.encoder.layer12.attention.qkv.weight\", kind: Parameter, value: ValueId(73), shape: TensorShape([1152, 3456]) }, GraphInput { name: \"vit.encoder.layer13.attention.qkv.weight\", kind: Parameter, value: ValueId(74), shape: TensorShape([1152, 3456]) }, GraphInput { name: \"vit.encoder.layer14.attention.qkv.weight\", kind: Parameter, value: ValueId(75), shape: TensorShape([1152, 3456]) }, GraphInput { name: \"vit.encoder.layer15.attention.qkv.weight\", kind: Parameter, value: ValueId(76), shape: TensorShape([1152, 3456]) }, GraphInput { name: \"vit.encoder.layer16.attention.qkv.weight\", kind: Parameter, value: ValueId(77), shape: TensorShape([1152, 3456]) }, GraphInput { name: \"vit.encoder.layer17.attention.qkv.weight\", kind: Parameter, value: ValueId(78), shape: TensorShape([1152, 3456]) }, GraphInput { name: \"vit.encoder.layer18.attention.qkv.weight\", kind: Parameter, value: ValueId(79), shape: TensorShape([1152, 3456]) }, GraphInput { name: \"vit.encoder.layer19.attention.qkv.weight\", kind: Parameter, value: ValueId(80), shape: TensorShape([1152, 3456]) }, GraphInput { name: \"vit.encoder.layer20.attention.qkv.weight\", kind: Parameter, value: ValueId(81), shape: TensorShape([1152, 3456]) }, GraphInput { name: \"vit.encoder.layer21.attention.qkv.weight\", kind: Parameter, value: ValueId(82), shape: TensorShape([1152, 3456]) }, GraphInput { name: \"vit.encoder.layer22.attention.qkv.weight\", kind: Parameter, value: ValueId(83), shape: TensorShape([1152, 3456]) }, GraphInput { name: \"vit.encoder.layer23.attention.qkv.weight\", kind: Parameter, value: ValueId(84), shape: TensorShape([1152, 3456]) }, GraphInput { name: \"vit.encoder.layer24.attention.qkv.weight\", kind: Parameter, value: ValueId(85), shape: TensorShape([1152, 3456]) }, GraphInput { name: \"vit.encoder.layer25.attention.qkv.weight\", kind: Parameter, value: ValueId(86), shape: TensorShape([1152, 3456]) }, GraphInput { name: \"vit.encoder.layer26.attention.qkv.weight\", kind: Parameter, value: ValueId(87), shape: TensorShape([1152, 3456]) }, GraphInput { name: \"vit.encoder.layer0.attention.qkv.bias\", kind: Parameter, value: ValueId(88), shape: TensorShape([1, 1, 3456]) }, GraphInput { name: \"vit.encoder.layer1.attention.qkv.bias\", kind: Parameter, value: ValueId(89), shape: TensorShape([1, 1, 3456]) }, GraphInput { name: \"vit.encoder.layer2.attention.qkv.bias\", kind: Parameter, value: ValueId(90), shape: TensorShape([1, 1, 3456]) }, GraphInput { name: \"vit.encoder.layer3.attention.qkv.bias\", kind: Parameter, value: ValueId(91), shape: TensorShape([1, 1, 3456]) }, GraphInput { name: \"vit.encoder.layer4.attention.qkv.bias\", kind: Parameter, value: ValueId(92), shape: TensorShape([1, 1, 3456]) }, GraphInput { name: \"vit.encoder.layer5.attention.qkv.bias\", kind: Parameter, value: ValueId(93), shape: TensorShape([1, 1, 3456]) }, GraphInput { name: \"vit.encoder.layer6.attention.qkv.bias\", kind: Parameter, value: ValueId(94), shape: TensorShape([1, 1, 3456]) }, GraphInput { name: \"vit.encoder.layer7.attention.qkv.bias\", kind: Parameter, value: ValueId(95), shape: TensorShape([1, 1, 3456]) }, GraphInput { name: \"vit.encoder.layer8.attention.qkv.bias\", kind: Parameter, value: ValueId(96), shape: TensorShape([1, 1, 3456]) }, GraphInput { name: \"vit.encoder.layer9.attention.qkv.bias\", kind: Parameter, value: ValueId(97), shape: TensorShape([1, 1, 3456]) }, GraphInput { name: \"vit.encoder.layer10.attention.qkv.bias\", kind: Parameter, value: ValueId(98), shape: TensorShape([1, 1, 3456]) }, GraphInput { name: \"vit.encoder.layer11.attention.qkv.bias\", kind: Parameter, value: ValueId(99), shape: TensorShape([1, 1, 3456]) }, GraphInput { name: \"vit.encoder.layer12.attention.qkv.bias\", kind: Parameter, value: ValueId(100), shape: TensorShape([1, 1, 3456]) }, GraphInput { name: \"vit.encoder.layer13.attention.qkv.bias\", kind: Parameter, value: ValueId(101), shape: TensorShape([1, 1, 3456]) }, GraphInput { name: \"vit.encoder.layer14.attention.qkv.bias\", kind: Parameter, value: ValueId(102), shape: TensorShape([1, 1, 3456]) }, GraphInput { name: \"vit.encoder.layer15.attention.qkv.bias\", kind: Parameter, value: ValueId(103), shape: TensorShape([1, 1, 3456]) }, GraphInput { name: \"vit.encoder.layer16.attention.qkv.bias\", kind: Parameter, value: ValueId(104), shape: TensorShape([1, 1, 3456]) }, GraphInput { name: \"vit.encoder.layer17.attention.qkv.bias\", kind: Parameter, value: ValueId(105), shape: TensorShape([1, 1, 3456]) }, GraphInput { name: \"vit.encoder.layer18.attention.qkv.bias\", kind: Parameter, value: ValueId(106), shape: TensorShape([1, 1, 3456]) }, GraphInput { name: \"vit.encoder.layer19.attention.qkv.bias\", kind: Parameter, value: ValueId(107), shape: TensorShape([1, 1, 3456]) }, GraphInput { name: \"vit.encoder.layer20.attention.qkv.bias\", kind: Parameter, value: ValueId(108), shape: TensorShape([1, 1, 3456]) }, GraphInput { name: \"vit.encoder.layer21.attention.qkv.bias\", kind: Parameter, value: ValueId(109), shape: TensorShape([1, 1, 3456]) }, GraphInput { name: \"vit.encoder.layer22.attention.qkv.bias\", kind: Parameter, value: ValueId(110), shape: TensorShape([1, 1, 3456]) }, GraphInput { name: \"vit.encoder.layer23.attention.qkv.bias\", kind: Parameter, value: ValueId(111), shape: TensorShape([1, 1, 3456]) }, GraphInput { name: \"vit.encoder.layer24.attention.qkv.bias\", kind: Parameter, value: ValueId(112), shape: TensorShape([1, 1, 3456]) }, GraphInput { name: \"vit.encoder.layer25.attention.qkv.bias\", kind: Parameter, value: ValueId(113), shape: TensorShape([1, 1, 3456]) }, GraphInput { name: \"vit.encoder.layer26.attention.qkv.bias\", kind: Parameter, value: ValueId(114), shape: TensorShape([1, 1, 3456]) }, GraphInput { name: \"vit.encoder.layer0.attention.output.weight\", kind: Parameter, value: ValueId(115), shape: TensorShape([1152, 1152]) }, GraphInput { name: \"vit.encoder.layer1.attention.output.weight\", kind: Parameter, value: ValueId(116), shape: TensorShape([1152, 1152]) }, GraphInput { name: \"vit.encoder.layer2.attention.output.weight\", kind: Parameter, value: ValueId(117), shape: TensorShape([1152, 1152]) }, GraphInput { name: \"vit.encoder.layer3.attention.output.weight\", kind: Parameter, value: ValueId(118), shape: TensorShape([1152, 1152]) }, GraphInput { name: \"vit.encoder.layer4.attention.output.weight\", kind: Parameter, value: ValueId(119), shape: TensorShape([1152, 1152]) }, GraphInput { name: \"vit.encoder.layer5.attention.output.weight\", kind: Parameter, value: ValueId(120), shape: TensorShape([1152, 1152]) }, GraphInput { name: \"vit.encoder.layer6.attention.output.weight\", kind: Parameter, value: ValueId(121), shape: TensorShape([1152, 1152]) }, GraphInput { name: \"vit.encoder.layer7.attention.output.weight\", kind: Parameter, value: ValueId(122), shape: TensorShape([1152, 1152]) }, GraphInput { name: \"vit.encoder.layer8.attention.output.weight\", kind: Parameter, value: ValueId(123), shape: TensorShape([1152, 1152]) }, GraphInput { name: \"vit.encoder.layer9.attention.output.weight\", kind: Parameter, value: ValueId(124), shape: TensorShape([1152, 1152]) }, GraphInput { name: \"vit.encoder.layer10.attention.output.weight\", kind: Parameter, value: ValueId(125), shape: TensorShape([1152, 1152]) }, GraphInput { name: \"vit.encoder.layer11.attention.output.weight\", kind: Parameter, value: ValueId(126), shape: TensorShape([1152, 1152]) }, GraphInput { name: \"vit.encoder.layer12.attention.output.weight\", kind: Parameter, value: ValueId(127), shape: TensorShape([1152, 1152]) }, GraphInput { name: \"vit.encoder.layer13.attention.output.weight\", kind: Parameter, value: ValueId(128), shape: TensorShape([1152, 1152]) }, GraphInput { name: \"vit.encoder.layer14.attention.output.weight\", kind: Parameter, value: ValueId(129), shape: TensorShape([1152, 1152]) }, GraphInput { name: \"vit.encoder.layer15.attention.output.weight\", kind: Parameter, value: ValueId(130), shape: TensorShape([1152, 1152]) }, GraphInput { name: \"vit.encoder.layer16.attention.output.weight\", kind: Parameter, value: ValueId(131), shape: TensorShape([1152, 1152]) }, GraphInput { name: \"vit.encoder.layer17.attention.output.weight\", kind: Parameter, value: ValueId(132), shape: TensorShape([1152, 1152]) }, GraphInput { name: \"vit.encoder.layer18.attention.output.weight\", kind: Parameter, value: ValueId(133), shape: TensorShape([1152, 1152]) }, GraphInput { name: \"vit.encoder.layer19.attention.output.weight\", kind: Parameter, value: ValueId(134), shape: TensorShape([1152, 1152]) }, GraphInput { name: \"vit.encoder.layer20.attention.output.weight\", kind: Parameter, value: ValueId(135), shape: TensorShape([1152, 1152]) }, GraphInput { name: \"vit.encoder.layer21.attention.output.weight\", kind: Parameter, value: ValueId(136), shape: TensorShape([1152, 1152]) }, GraphInput { name: \"vit.encoder.layer22.attention.output.weight\", kind: Parameter, value: ValueId(137), shape: TensorShape([1152, 1152]) }, GraphInput { name: \"vit.encoder.layer23.attention.output.weight\", kind: Parameter, value: ValueId(138), shape: TensorShape([1152, 1152]) }, GraphInput { name: \"vit.encoder.layer24.attention.output.weight\", kind: Parameter, value: ValueId(139), shape: TensorShape([1152, 1152]) }, GraphInput { name: \"vit.encoder.layer25.attention.output.weight\", kind: Parameter, value: ValueId(140), shape: TensorShape([1152, 1152]) }, GraphInput { name: \"vit.encoder.layer26.attention.output.weight\", kind: Parameter, value: ValueId(141), shape: TensorShape([1152, 1152]) }, GraphInput { name: \"vit.encoder.layer0.attention.output.bias\", kind: Parameter, value: ValueId(142), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer1.attention.output.bias\", kind: Parameter, value: ValueId(143), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer2.attention.output.bias\", kind: Parameter, value: ValueId(144), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer3.attention.output.bias\", kind: Parameter, value: ValueId(145), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer4.attention.output.bias\", kind: Parameter, value: ValueId(146), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer5.attention.output.bias\", kind: Parameter, value: ValueId(147), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer6.attention.output.bias\", kind: Parameter, value: ValueId(148), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer7.attention.output.bias\", kind: Parameter, value: ValueId(149), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer8.attention.output.bias\", kind: Parameter, value: ValueId(150), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer9.attention.output.bias\", kind: Parameter, value: ValueId(151), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer10.attention.output.bias\", kind: Parameter, value: ValueId(152), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer11.attention.output.bias\", kind: Parameter, value: ValueId(153), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer12.attention.output.bias\", kind: Parameter, value: ValueId(154), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer13.attention.output.bias\", kind: Parameter, value: ValueId(155), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer14.attention.output.bias\", kind: Parameter, value: ValueId(156), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer15.attention.output.bias\", kind: Parameter, value: ValueId(157), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer16.attention.output.bias\", kind: Parameter, value: ValueId(158), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer17.attention.output.bias\", kind: Parameter, value: ValueId(159), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer18.attention.output.bias\", kind: Parameter, value: ValueId(160), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer19.attention.output.bias\", kind: Parameter, value: ValueId(161), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer20.attention.output.bias\", kind: Parameter, value: ValueId(162), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer21.attention.output.bias\", kind: Parameter, value: ValueId(163), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer22.attention.output.bias\", kind: Parameter, value: ValueId(164), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer23.attention.output.bias\", kind: Parameter, value: ValueId(165), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer24.attention.output.bias\", kind: Parameter, value: ValueId(166), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer25.attention.output.bias\", kind: Parameter, value: ValueId(167), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer26.attention.output.bias\", kind: Parameter, value: ValueId(168), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer0.mlp_norm.scale\", kind: Parameter, value: ValueId(169), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer1.mlp_norm.scale\", kind: Parameter, value: ValueId(170), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer2.mlp_norm.scale\", kind: Parameter, value: ValueId(171), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer3.mlp_norm.scale\", kind: Parameter, value: ValueId(172), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer4.mlp_norm.scale\", kind: Parameter, value: ValueId(173), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer5.mlp_norm.scale\", kind: Parameter, value: ValueId(174), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer6.mlp_norm.scale\", kind: Parameter, value: ValueId(175), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer7.mlp_norm.scale\", kind: Parameter, value: ValueId(176), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer8.mlp_norm.scale\", kind: Parameter, value: ValueId(177), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer9.mlp_norm.scale\", kind: Parameter, value: ValueId(178), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer10.mlp_norm.scale\", kind: Parameter, value: ValueId(179), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer11.mlp_norm.scale\", kind: Parameter, value: ValueId(180), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer12.mlp_norm.scale\", kind: Parameter, value: ValueId(181), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer13.mlp_norm.scale\", kind: Parameter, value: ValueId(182), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer14.mlp_norm.scale\", kind: Parameter, value: ValueId(183), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer15.mlp_norm.scale\", kind: Parameter, value: ValueId(184), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer16.mlp_norm.scale\", kind: Parameter, value: ValueId(185), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer17.mlp_norm.scale\", kind: Parameter, value: ValueId(186), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer18.mlp_norm.scale\", kind: Parameter, value: ValueId(187), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer19.mlp_norm.scale\", kind: Parameter, value: ValueId(188), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer20.mlp_norm.scale\", kind: Parameter, value: ValueId(189), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer21.mlp_norm.scale\", kind: Parameter, value: ValueId(190), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer22.mlp_norm.scale\", kind: Parameter, value: ValueId(191), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer23.mlp_norm.scale\", kind: Parameter, value: ValueId(192), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer24.mlp_norm.scale\", kind: Parameter, value: ValueId(193), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer25.mlp_norm.scale\", kind: Parameter, value: ValueId(194), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer26.mlp_norm.scale\", kind: Parameter, value: ValueId(195), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer0.mlp_norm.bias\", kind: Parameter, value: ValueId(196), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer1.mlp_norm.bias\", kind: Parameter, value: ValueId(197), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer2.mlp_norm.bias\", kind: Parameter, value: ValueId(198), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer3.mlp_norm.bias\", kind: Parameter, value: ValueId(199), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer4.mlp_norm.bias\", kind: Parameter, value: ValueId(200), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer5.mlp_norm.bias\", kind: Parameter, value: ValueId(201), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer6.mlp_norm.bias\", kind: Parameter, value: ValueId(202), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer7.mlp_norm.bias\", kind: Parameter, value: ValueId(203), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer8.mlp_norm.bias\", kind: Parameter, value: ValueId(204), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer9.mlp_norm.bias\", kind: Parameter, value: ValueId(205), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer10.mlp_norm.bias\", kind: Parameter, value: ValueId(206), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer11.mlp_norm.bias\", kind: Parameter, value: ValueId(207), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer12.mlp_norm.bias\", kind: Parameter, value: ValueId(208), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer13.mlp_norm.bias\", kind: Parameter, value: ValueId(209), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer14.mlp_norm.bias\", kind: Parameter, value: ValueId(210), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer15.mlp_norm.bias\", kind: Parameter, value: ValueId(211), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer16.mlp_norm.bias\", kind: Parameter, value: ValueId(212), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer17.mlp_norm.bias\", kind: Parameter, value: ValueId(213), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer18.mlp_norm.bias\", kind: Parameter, value: ValueId(214), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer19.mlp_norm.bias\", kind: Parameter, value: ValueId(215), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer20.mlp_norm.bias\", kind: Parameter, value: ValueId(216), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer21.mlp_norm.bias\", kind: Parameter, value: ValueId(217), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer22.mlp_norm.bias\", kind: Parameter, value: ValueId(218), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer23.mlp_norm.bias\", kind: Parameter, value: ValueId(219), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer24.mlp_norm.bias\", kind: Parameter, value: ValueId(220), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer25.mlp_norm.bias\", kind: Parameter, value: ValueId(221), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer26.mlp_norm.bias\", kind: Parameter, value: ValueId(222), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer0.mlp.up.weight\", kind: Parameter, value: ValueId(223), shape: TensorShape([1152, 4304]) }, GraphInput { name: \"vit.encoder.layer1.mlp.up.weight\", kind: Parameter, value: ValueId(224), shape: TensorShape([1152, 4304]) }, GraphInput { name: \"vit.encoder.layer2.mlp.up.weight\", kind: Parameter, value: ValueId(225), shape: TensorShape([1152, 4304]) }, GraphInput { name: \"vit.encoder.layer3.mlp.up.weight\", kind: Parameter, value: ValueId(226), shape: TensorShape([1152, 4304]) }, GraphInput { name: \"vit.encoder.layer4.mlp.up.weight\", kind: Parameter, value: ValueId(227), shape: TensorShape([1152, 4304]) }, GraphInput { name: \"vit.encoder.layer5.mlp.up.weight\", kind: Parameter, value: ValueId(228), shape: TensorShape([1152, 4304]) }, GraphInput { name: \"vit.encoder.layer6.mlp.up.weight\", kind: Parameter, value: ValueId(229), shape: TensorShape([1152, 4304]) }, GraphInput { name: \"vit.encoder.layer7.mlp.up.weight\", kind: Parameter, value: ValueId(230), shape: TensorShape([1152, 4304]) }, GraphInput { name: \"vit.encoder.layer8.mlp.up.weight\", kind: Parameter, value: ValueId(231), shape: TensorShape([1152, 4304]) }, GraphInput { name: \"vit.encoder.layer9.mlp.up.weight\", kind: Parameter, value: ValueId(232), shape: TensorShape([1152, 4304]) }, GraphInput { name: \"vit.encoder.layer10.mlp.up.weight\", kind: Parameter, value: ValueId(233), shape: TensorShape([1152, 4304]) }, GraphInput { name: \"vit.encoder.layer11.mlp.up.weight\", kind: Parameter, value: ValueId(234), shape: TensorShape([1152, 4304]) }, GraphInput { name: \"vit.encoder.layer12.mlp.up.weight\", kind: Parameter, value: ValueId(235), shape: TensorShape([1152, 4304]) }, GraphInput { name: \"vit.encoder.layer13.mlp.up.weight\", kind: Parameter, value: ValueId(236), shape: TensorShape([1152, 4304]) }, GraphInput { name: \"vit.encoder.layer14.mlp.up.weight\", kind: Parameter, value: ValueId(237), shape: TensorShape([1152, 4304]) }, GraphInput { name: \"vit.encoder.layer15.mlp.up.weight\", kind: Parameter, value: ValueId(238), shape: TensorShape([1152, 4304]) }, GraphInput { name: \"vit.encoder.layer16.mlp.up.weight\", kind: Parameter, value: ValueId(239), shape: TensorShape([1152, 4304]) }, GraphInput { name: \"vit.encoder.layer17.mlp.up.weight\", kind: Parameter, value: ValueId(240), shape: TensorShape([1152, 4304]) }, GraphInput { name: \"vit.encoder.layer18.mlp.up.weight\", kind: Parameter, value: ValueId(241), shape: TensorShape([1152, 4304]) }, GraphInput { name: \"vit.encoder.layer19.mlp.up.weight\", kind: Parameter, value: ValueId(242), shape: TensorShape([1152, 4304]) }, GraphInput { name: \"vit.encoder.layer20.mlp.up.weight\", kind: Parameter, value: ValueId(243), shape: TensorShape([1152, 4304]) }, GraphInput { name: \"vit.encoder.layer21.mlp.up.weight\", kind: Parameter, value: ValueId(244), shape: TensorShape([1152, 4304]) }, GraphInput { name: \"vit.encoder.layer22.mlp.up.weight\", kind: Parameter, value: ValueId(245), shape: TensorShape([1152, 4304]) }, GraphInput { name: \"vit.encoder.layer23.mlp.up.weight\", kind: Parameter, value: ValueId(246), shape: TensorShape([1152, 4304]) }, GraphInput { name: \"vit.encoder.layer24.mlp.up.weight\", kind: Parameter, value: ValueId(247), shape: TensorShape([1152, 4304]) }, GraphInput { name: \"vit.encoder.layer25.mlp.up.weight\", kind: Parameter, value: ValueId(248), shape: TensorShape([1152, 4304]) }, GraphInput { name: \"vit.encoder.layer26.mlp.up.weight\", kind: Parameter, value: ValueId(249), shape: TensorShape([1152, 4304]) }, GraphInput { name: \"vit.encoder.layer0.mlp.up.bias\", kind: Parameter, value: ValueId(250), shape: TensorShape([1, 1, 4304]) }, GraphInput { name: \"vit.encoder.layer1.mlp.up.bias\", kind: Parameter, value: ValueId(251), shape: TensorShape([1, 1, 4304]) }, GraphInput { name: \"vit.encoder.layer2.mlp.up.bias\", kind: Parameter, value: ValueId(252), shape: TensorShape([1, 1, 4304]) }, GraphInput { name: \"vit.encoder.layer3.mlp.up.bias\", kind: Parameter, value: ValueId(253), shape: TensorShape([1, 1, 4304]) }, GraphInput { name: \"vit.encoder.layer4.mlp.up.bias\", kind: Parameter, value: ValueId(254), shape: TensorShape([1, 1, 4304]) }, GraphInput { name: \"vit.encoder.layer5.mlp.up.bias\", kind: Parameter, value: ValueId(255), shape: TensorShape([1, 1, 4304]) }, GraphInput { name: \"vit.encoder.layer6.mlp.up.bias\", kind: Parameter, value: ValueId(256), shape: TensorShape([1, 1, 4304]) }, GraphInput { name: \"vit.encoder.layer7.mlp.up.bias\", kind: Parameter, value: ValueId(257), shape: TensorShape([1, 1, 4304]) }, GraphInput { name: \"vit.encoder.layer8.mlp.up.bias\", kind: Parameter, value: ValueId(258), shape: TensorShape([1, 1, 4304]) }, GraphInput { name: \"vit.encoder.layer9.mlp.up.bias\", kind: Parameter, value: ValueId(259), shape: TensorShape([1, 1, 4304]) }, GraphInput { name: \"vit.encoder.layer10.mlp.up.bias\", kind: Parameter, value: ValueId(260), shape: TensorShape([1, 1, 4304]) }, GraphInput { name: \"vit.encoder.layer11.mlp.up.bias\", kind: Parameter, value: ValueId(261), shape: TensorShape([1, 1, 4304]) }, GraphInput { name: \"vit.encoder.layer12.mlp.up.bias\", kind: Parameter, value: ValueId(262), shape: TensorShape([1, 1, 4304]) }, GraphInput { name: \"vit.encoder.layer13.mlp.up.bias\", kind: Parameter, value: ValueId(263), shape: TensorShape([1, 1, 4304]) }, GraphInput { name: \"vit.encoder.layer14.mlp.up.bias\", kind: Parameter, value: ValueId(264), shape: TensorShape([1, 1, 4304]) }, GraphInput { name: \"vit.encoder.layer15.mlp.up.bias\", kind: Parameter, value: ValueId(265), shape: TensorShape([1, 1, 4304]) }, GraphInput { name: \"vit.encoder.layer16.mlp.up.bias\", kind: Parameter, value: ValueId(266), shape: TensorShape([1, 1, 4304]) }, GraphInput { name: \"vit.encoder.layer17.mlp.up.bias\", kind: Parameter, value: ValueId(267), shape: TensorShape([1, 1, 4304]) }, GraphInput { name: \"vit.encoder.layer18.mlp.up.bias\", kind: Parameter, value: ValueId(268), shape: TensorShape([1, 1, 4304]) }, GraphInput { name: \"vit.encoder.layer19.mlp.up.bias\", kind: Parameter, value: ValueId(269), shape: TensorShape([1, 1, 4304]) }, GraphInput { name: \"vit.encoder.layer20.mlp.up.bias\", kind: Parameter, value: ValueId(270), shape: TensorShape([1, 1, 4304]) }, GraphInput { name: \"vit.encoder.layer21.mlp.up.bias\", kind: Parameter, value: ValueId(271), shape: TensorShape([1, 1, 4304]) }, GraphInput { name: \"vit.encoder.layer22.mlp.up.bias\", kind: Parameter, value: ValueId(272), shape: TensorShape([1, 1, 4304]) }, GraphInput { name: \"vit.encoder.layer23.mlp.up.bias\", kind: Parameter, value: ValueId(273), shape: TensorShape([1, 1, 4304]) }, GraphInput { name: \"vit.encoder.layer24.mlp.up.bias\", kind: Parameter, value: ValueId(274), shape: TensorShape([1, 1, 4304]) }, GraphInput { name: \"vit.encoder.layer25.mlp.up.bias\", kind: Parameter, value: ValueId(275), shape: TensorShape([1, 1, 4304]) }, GraphInput { name: \"vit.encoder.layer26.mlp.up.bias\", kind: Parameter, value: ValueId(276), shape: TensorShape([1, 1, 4304]) }, GraphInput { name: \"vit.encoder.layer0.mlp.down.weight\", kind: Parameter, value: ValueId(277), shape: TensorShape([4304, 1152]) }, GraphInput { name: \"vit.encoder.layer1.mlp.down.weight\", kind: Parameter, value: ValueId(278), shape: TensorShape([4304, 1152]) }, GraphInput { name: \"vit.encoder.layer2.mlp.down.weight\", kind: Parameter, value: ValueId(279), shape: TensorShape([4304, 1152]) }, GraphInput { name: \"vit.encoder.layer3.mlp.down.weight\", kind: Parameter, value: ValueId(280), shape: TensorShape([4304, 1152]) }, GraphInput { name: \"vit.encoder.layer4.mlp.down.weight\", kind: Parameter, value: ValueId(281), shape: TensorShape([4304, 1152]) }, GraphInput { name: \"vit.encoder.layer5.mlp.down.weight\", kind: Parameter, value: ValueId(282), shape: TensorShape([4304, 1152]) }, GraphInput { name: \"vit.encoder.layer6.mlp.down.weight\", kind: Parameter, value: ValueId(283), shape: TensorShape([4304, 1152]) }, GraphInput { name: \"vit.encoder.layer7.mlp.down.weight\", kind: Parameter, value: ValueId(284), shape: TensorShape([4304, 1152]) }, GraphInput { name: \"vit.encoder.layer8.mlp.down.weight\", kind: Parameter, value: ValueId(285), shape: TensorShape([4304, 1152]) }, GraphInput { name: \"vit.encoder.layer9.mlp.down.weight\", kind: Parameter, value: ValueId(286), shape: TensorShape([4304, 1152]) }, GraphInput { name: \"vit.encoder.layer10.mlp.down.weight\", kind: Parameter, value: ValueId(287), shape: TensorShape([4304, 1152]) }, GraphInput { name: \"vit.encoder.layer11.mlp.down.weight\", kind: Parameter, value: ValueId(288), shape: TensorShape([4304, 1152]) }, GraphInput { name: \"vit.encoder.layer12.mlp.down.weight\", kind: Parameter, value: ValueId(289), shape: TensorShape([4304, 1152]) }, GraphInput { name: \"vit.encoder.layer13.mlp.down.weight\", kind: Parameter, value: ValueId(290), shape: TensorShape([4304, 1152]) }, GraphInput { name: \"vit.encoder.layer14.mlp.down.weight\", kind: Parameter, value: ValueId(291), shape: TensorShape([4304, 1152]) }, GraphInput { name: \"vit.encoder.layer15.mlp.down.weight\", kind: Parameter, value: ValueId(292), shape: TensorShape([4304, 1152]) }, GraphInput { name: \"vit.encoder.layer16.mlp.down.weight\", kind: Parameter, value: ValueId(293), shape: TensorShape([4304, 1152]) }, GraphInput { name: \"vit.encoder.layer17.mlp.down.weight\", kind: Parameter, value: ValueId(294), shape: TensorShape([4304, 1152]) }, GraphInput { name: \"vit.encoder.layer18.mlp.down.weight\", kind: Parameter, value: ValueId(295), shape: TensorShape([4304, 1152]) }, GraphInput { name: \"vit.encoder.layer19.mlp.down.weight\", kind: Parameter, value: ValueId(296), shape: TensorShape([4304, 1152]) }, GraphInput { name: \"vit.encoder.layer20.mlp.down.weight\", kind: Parameter, value: ValueId(297), shape: TensorShape([4304, 1152]) }, GraphInput { name: \"vit.encoder.layer21.mlp.down.weight\", kind: Parameter, value: ValueId(298), shape: TensorShape([4304, 1152]) }, GraphInput { name: \"vit.encoder.layer22.mlp.down.weight\", kind: Parameter, value: ValueId(299), shape: TensorShape([4304, 1152]) }, GraphInput { name: \"vit.encoder.layer23.mlp.down.weight\", kind: Parameter, value: ValueId(300), shape: TensorShape([4304, 1152]) }, GraphInput { name: \"vit.encoder.layer24.mlp.down.weight\", kind: Parameter, value: ValueId(301), shape: TensorShape([4304, 1152]) }, GraphInput { name: \"vit.encoder.layer25.mlp.down.weight\", kind: Parameter, value: ValueId(302), shape: TensorShape([4304, 1152]) }, GraphInput { name: \"vit.encoder.layer26.mlp.down.weight\", kind: Parameter, value: ValueId(303), shape: TensorShape([4304, 1152]) }, GraphInput { name: \"vit.encoder.layer0.mlp.down.bias\", kind: Parameter, value: ValueId(304), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer1.mlp.down.bias\", kind: Parameter, value: ValueId(305), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer2.mlp.down.bias\", kind: Parameter, value: ValueId(306), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer3.mlp.down.bias\", kind: Parameter, value: ValueId(307), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer4.mlp.down.bias\", kind: Parameter, value: ValueId(308), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer5.mlp.down.bias\", kind: Parameter, value: ValueId(309), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer6.mlp.down.bias\", kind: Parameter, value: ValueId(310), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer7.mlp.down.bias\", kind: Parameter, value: ValueId(311), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer8.mlp.down.bias\", kind: Parameter, value: ValueId(312), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer9.mlp.down.bias\", kind: Parameter, value: ValueId(313), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer10.mlp.down.bias\", kind: Parameter, value: ValueId(314), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer11.mlp.down.bias\", kind: Parameter, value: ValueId(315), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer12.mlp.down.bias\", kind: Parameter, value: ValueId(316), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer13.mlp.down.bias\", kind: Parameter, value: ValueId(317), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer14.mlp.down.bias\", kind: Parameter, value: ValueId(318), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer15.mlp.down.bias\", kind: Parameter, value: ValueId(319), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer16.mlp.down.bias\", kind: Parameter, value: ValueId(320), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer17.mlp.down.bias\", kind: Parameter, value: ValueId(321), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer18.mlp.down.bias\", kind: Parameter, value: ValueId(322), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer19.mlp.down.bias\", kind: Parameter, value: ValueId(323), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer20.mlp.down.bias\", kind: Parameter, value: ValueId(324), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer21.mlp.down.bias\", kind: Parameter, value: ValueId(325), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer22.mlp.down.bias\", kind: Parameter, value: ValueId(326), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer23.mlp.down.bias\", kind: Parameter, value: ValueId(327), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer24.mlp.down.bias\", kind: Parameter, value: ValueId(328), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer25.mlp.down.bias\", kind: Parameter, value: ValueId(329), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.layer26.mlp.down.bias\", kind: Parameter, value: ValueId(330), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.final_norm.scale\", kind: Parameter, value: ValueId(366), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.encoder.final_norm.bias\", kind: Parameter, value: ValueId(367), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.map.probe\", kind: Parameter, value: ValueId(369), shape: TensorShape([2, 1, 1152]) }, GraphInput { name: \"vit.map.attention.kv.weight\", kind: Parameter, value: ValueId(370), shape: TensorShape([1152, 2304]) }, GraphInput { name: \"vit.map.attention.kv.bias\", kind: Parameter, value: ValueId(371), shape: TensorShape([1, 1, 2304]) }, GraphInput { name: \"vit.map.attention.query.weight\", kind: Parameter, value: ValueId(374), shape: TensorShape([1152, 1152]) }, GraphInput { name: \"vit.map.attention.query.bias\", kind: Parameter, value: ValueId(375), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.map.attention.output.weight\", kind: Parameter, value: ValueId(385), shape: TensorShape([1152, 1152]) }, GraphInput { name: \"vit.map.attention.output.bias\", kind: Parameter, value: ValueId(386), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.map.norm.scale\", kind: Parameter, value: ValueId(389), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.map.norm.bias\", kind: Parameter, value: ValueId(390), shape: TensorShape([1, 1, 1152]) }, GraphInput { name: \"vit.map.mlp.up.weight\", kind: Parameter, value: ValueId(392), shape: TensorShape([1152, 4304]) }, GraphInput { name: \"vit.map.mlp.up.bias\", kind: Parameter, value: ValueId(393), shape: TensorShape([1, 1, 4304]) }, GraphInput { name: \"vit.map.mlp.down.weight\", kind: Parameter, value: ValueId(397), shape: TensorShape([4304, 1152]) }, GraphInput { name: \"vit.map.mlp.down.bias\", kind: Parameter, value: ValueId(398), shape: TensorShape([1, 1, 1152]) }], sequences: [ValueSequence { id: ValueSequenceId(0), name: \"vit.encoder.attention_norm.scale\", values: [ValueId(7), ValueId(8), ValueId(9), ValueId(10), ValueId(11), ValueId(12), ValueId(13), ValueId(14), ValueId(15), ValueId(16), ValueId(17), ValueId(18), ValueId(19), ValueId(20), ValueId(21), ValueId(22), ValueId(23), ValueId(24), ValueId(25), ValueId(26), ValueId(27), ValueId(28), ValueId(29), ValueId(30), ValueId(31), ValueId(32), ValueId(33)] }, ValueSequence { id: ValueSequenceId(1), name: \"vit.encoder.attention_norm.bias\", values: [ValueId(34), ValueId(35), ValueId(36), ValueId(37), ValueId(38), ValueId(39), ValueId(40), ValueId(41), ValueId(42), ValueId(43), ValueId(44), ValueId(45), ValueId(46), ValueId(47), ValueId(48), ValueId(49), ValueId(50), ValueId(51), ValueId(52), ValueId(53), ValueId(54), ValueId(55), ValueId(56), ValueId(57), ValueId(58), ValueId(59), ValueId(60)] }, ValueSequence { id: ValueSequenceId(2), name: \"vit.encoder.attention.qkv.weight\", values: [ValueId(61), ValueId(62), ValueId(63), ValueId(64), ValueId(65), ValueId(66), ValueId(67), ValueId(68), ValueId(69), ValueId(70), ValueId(71), ValueId(72), ValueId(73), ValueId(74), ValueId(75), ValueId(76), ValueId(77), ValueId(78), ValueId(79), ValueId(80), ValueId(81), ValueId(82), ValueId(83), ValueId(84), ValueId(85), ValueId(86), ValueId(87)] }, ValueSequence { id: ValueSequenceId(3), name: \"vit.encoder.attention.qkv.bias\", values: [ValueId(88), ValueId(89), ValueId(90), ValueId(91), ValueId(92), ValueId(93), ValueId(94), ValueId(95), ValueId(96), ValueId(97), ValueId(98), ValueId(99), ValueId(100), ValueId(101), ValueId(102), ValueId(103), ValueId(104), ValueId(105), ValueId(106), ValueId(107), ValueId(108), ValueId(109), ValueId(110), ValueId(111), ValueId(112), ValueId(113), ValueId(114)] }, ValueSequence { id: ValueSequenceId(4), name: \"vit.encoder.attention.output.weight\", values: [ValueId(115), ValueId(116), ValueId(117), ValueId(118), ValueId(119), ValueId(120), ValueId(121), ValueId(122), ValueId(123), ValueId(124), ValueId(125), ValueId(126), ValueId(127), ValueId(128), ValueId(129), ValueId(130), ValueId(131), ValueId(132), ValueId(133), ValueId(134), ValueId(135), ValueId(136), ValueId(137), ValueId(138), ValueId(139), ValueId(140), ValueId(141)] }, ValueSequence { id: ValueSequenceId(5), name: \"vit.encoder.attention.output.bias\", values: [ValueId(142), ValueId(143), ValueId(144), ValueId(145), ValueId(146), ValueId(147), ValueId(148), ValueId(149), ValueId(150), ValueId(151), ValueId(152), ValueId(153), ValueId(154), ValueId(155), ValueId(156), ValueId(157), ValueId(158), ValueId(159), ValueId(160), ValueId(161), ValueId(162), ValueId(163), ValueId(164), ValueId(165), ValueId(166), ValueId(167), ValueId(168)] }, ValueSequence { id: ValueSequenceId(6), name: \"vit.encoder.mlp_norm.scale\", values: [ValueId(169), ValueId(170), ValueId(171), ValueId(172), ValueId(173), ValueId(174), ValueId(175), ValueId(176), ValueId(177), ValueId(178), ValueId(179), ValueId(180), ValueId(181), ValueId(182), ValueId(183), ValueId(184), ValueId(185), ValueId(186), ValueId(187), ValueId(188), ValueId(189), ValueId(190), ValueId(191), ValueId(192), ValueId(193), ValueId(194), ValueId(195)] }, ValueSequence { id: ValueSequenceId(7), name: \"vit.encoder.mlp_norm.bias\", values: [ValueId(196), ValueId(197), ValueId(198), ValueId(199), ValueId(200), ValueId(201), ValueId(202), ValueId(203), ValueId(204), ValueId(205), ValueId(206), ValueId(207), ValueId(208), ValueId(209), ValueId(210), ValueId(211), ValueId(212), ValueId(213), ValueId(214), ValueId(215), ValueId(216), ValueId(217), ValueId(218), ValueId(219), ValueId(220), ValueId(221), ValueId(222)] }, ValueSequence { id: ValueSequenceId(8), name: \"vit.encoder.mlp.up.weight\", values: [ValueId(223), ValueId(224), ValueId(225), ValueId(226), ValueId(227), ValueId(228), ValueId(229), ValueId(230), ValueId(231), ValueId(232), ValueId(233), ValueId(234), ValueId(235), ValueId(236), ValueId(237), ValueId(238), ValueId(239), ValueId(240), ValueId(241), ValueId(242), ValueId(243), ValueId(244), ValueId(245), ValueId(246), ValueId(247), ValueId(248), ValueId(249)] }, ValueSequence { id: ValueSequenceId(9), name: \"vit.encoder.mlp.up.bias\", values: [ValueId(250), ValueId(251), ValueId(252), ValueId(253), ValueId(254), ValueId(255), ValueId(256), ValueId(257), ValueId(258), ValueId(259), ValueId(260), ValueId(261), ValueId(262), ValueId(263), ValueId(264), ValueId(265), ValueId(266), ValueId(267), ValueId(268), ValueId(269), ValueId(270), ValueId(271), ValueId(272), ValueId(273), ValueId(274), ValueId(275), ValueId(276)] }, ValueSequence { id: ValueSequenceId(10), name: \"vit.encoder.mlp.down.weight\", values: [ValueId(277), ValueId(278), ValueId(279), ValueId(280), ValueId(281), ValueId(282), ValueId(283), ValueId(284), ValueId(285), ValueId(286), ValueId(287), ValueId(288), ValueId(289), ValueId(290), ValueId(291), ValueId(292), ValueId(293), ValueId(294), ValueId(295), ValueId(296), ValueId(297), ValueId(298), ValueId(299), ValueId(300), ValueId(301), ValueId(302), ValueId(303)] }, ValueSequence { id: ValueSequenceId(11), name: \"vit.encoder.mlp.down.bias\", values: [ValueId(304), ValueId(305), ValueId(306), ValueId(307), ValueId(308), ValueId(309), ValueId(310), ValueId(311), ValueId(312), ValueId(313), ValueId(314), ValueId(315), ValueId(316), ValueId(317), ValueId(318), ValueId(319), ValueId(320), ValueId(321), ValueId(322), ValueId(323), ValueId(324), ValueId(325), ValueId(326), ValueId(327), ValueId(328), ValueId(329), ValueId(330)] }], operations: [Operation { id: OperationId(0), inputs: [ValueId(0), ValueId(1)], results: [ValueId(3)], kind: Gemm(GemmOptions { transpose_left: false, transpose_right: false }) }, Operation { id: OperationId(1), inputs: [ValueId(3), ValueId(2)], results: [ValueId(4)], kind: Add }, Operation { id: OperationId(2), inputs: [ValueId(4), ValueId(5)], results: [ValueId(6)], kind: Add }, Operation { id: OperationId(24), inputs: [ValueId(6)], results: [ValueId(365)], kind: Repeat(Repeat { count: 27, carried_inputs: 1, invariant_inputs: 0, iterated_inputs: [ValueSequenceId(0), ValueSequenceId(1), ValueSequenceId(2), ValueSequenceId(3), ValueSequenceId(4), ValueSequenceId(5), ValueSequenceId(6), ValueSequenceId(7), ValueSequenceId(8), ValueSequenceId(9), ValueSequenceId(10), ValueSequenceId(11)], body: Region { arguments: [ValueId(331), ValueId(332), ValueId(333), ValueId(334), ValueId(335), ValueId(336), ValueId(337), ValueId(338), ValueId(339), ValueId(340), ValueId(341), ValueId(342), ValueId(343)], operations: [Operation { id: OperationId(3), inputs: [ValueId(331), ValueId(332), ValueId(333)], results: [ValueId(344)], kind: LayerNorm }, Operation { id: OperationId(4), inputs: [ValueId(344), ValueId(334)], results: [ValueId(345)], kind: Gemm(GemmOptions { transpose_left: false, transpose_right: false }) }, Operation { id: OperationId(5), inputs: [ValueId(345), ValueId(335)], results: [ValueId(346)], kind: Add }, Operation { id: OperationId(6), inputs: [ValueId(346)], results: [ValueId(347)], kind: Slice(AxisSlice { axis: 2, start: 0, length: 1152 }) }, Operation { id: OperationId(7), inputs: [ValueId(346)], results: [ValueId(348)], kind: Slice(AxisSlice { axis: 2, start: 1152, length: 1152 }) }, Operation { id: OperationId(8), inputs: [ValueId(346)], results: [ValueId(349)], kind: Slice(AxisSlice { axis: 2, start: 2304, length: 1152 }) }, Operation { id: OperationId(9), inputs: [ValueId(347)], results: [ValueId(350)], kind: View(AxisFactorView { split_axis: 2, merge_axis: 0, factor: 16, reversed: false }) }, Operation { id: OperationId(10), inputs: [ValueId(348)], results: [ValueId(351)], kind: View(AxisFactorView { split_axis: 2, merge_axis: 0, factor: 16, reversed: false }) }, Operation { id: OperationId(11), inputs: [ValueId(349)], results: [ValueId(352)], kind: View(AxisFactorView { split_axis: 2, merge_axis: 0, factor: 16, reversed: false }) }, Operation { id: OperationId(12), inputs: [ValueId(350), ValueId(351), ValueId(352)], results: [ValueId(353)], kind: FlashAttention(AttentionOptions { causal: false, scale: InverseSqrtQueryWidth }) }, Operation { id: OperationId(13), inputs: [ValueId(353)], results: [ValueId(354)], kind: View(AxisFactorView { split_axis: 2, merge_axis: 0, factor: 16, reversed: true }) }, Operation { id: OperationId(14), inputs: [ValueId(354), ValueId(336)], results: [ValueId(355)], kind: Gemm(GemmOptions { transpose_left: false, transpose_right: false }) }, Operation { id: OperationId(15), inputs: [ValueId(355), ValueId(337)], results: [ValueId(356)], kind: Add }, Operation { id: OperationId(16), inputs: [ValueId(331), ValueId(356)], results: [ValueId(357)], kind: Add }, Operation { id: OperationId(17), inputs: [ValueId(357), ValueId(338), ValueId(339)], results: [ValueId(358)], kind: LayerNorm }, Operation { id: OperationId(18), inputs: [ValueId(358), ValueId(340)], results: [ValueId(359)], kind: Gemm(GemmOptions { transpose_left: false, transpose_right: false }) }, Operation { id: OperationId(19), inputs: [ValueId(359), ValueId(341)], results: [ValueId(360)], kind: Add }, Operation { id: OperationId(20), inputs: [ValueId(360)], results: [ValueId(361)], kind: Gelu }, Operation { id: OperationId(21), inputs: [ValueId(361), ValueId(342)], results: [ValueId(362)], kind: Gemm(GemmOptions { transpose_left: false, transpose_right: false }) }, Operation { id: OperationId(22), inputs: [ValueId(362), ValueId(343)], results: [ValueId(363)], kind: Add }, Operation { id: OperationId(23), inputs: [ValueId(357), ValueId(363)], results: [ValueId(364)], kind: Add }], yields: [ValueId(364)], value_shapes: {ValueId(331): TensorShape([2, 729, 1152]), ValueId(332): TensorShape([1, 1, 1152]), ValueId(333): TensorShape([1, 1, 1152]), ValueId(334): TensorShape([1152, 3456]), ValueId(335): TensorShape([1, 1, 3456]), ValueId(336): TensorShape([1152, 1152]), ValueId(337): TensorShape([1, 1, 1152]), ValueId(338): TensorShape([1, 1, 1152]), ValueId(339): TensorShape([1, 1, 1152]), ValueId(340): TensorShape([1152, 4304]), ValueId(341): TensorShape([1, 1, 4304]), ValueId(342): TensorShape([4304, 1152]), ValueId(343): TensorShape([1, 1, 1152]), ValueId(344): TensorShape([2, 729, 1152]), ValueId(345): TensorShape([2, 729, 3456]), ValueId(346): TensorShape([2, 729, 3456]), ValueId(347): TensorShape([2, 729, 1152]), ValueId(348): TensorShape([2, 729, 1152]), ValueId(349): TensorShape([2, 729, 1152]), ValueId(350): TensorShape([32, 729, 72]), ValueId(351): TensorShape([32, 729, 72]), ValueId(352): TensorShape([32, 729, 72]), ValueId(353): TensorShape([32, 729, 72]), ValueId(354): TensorShape([2, 729, 1152]), ValueId(355): TensorShape([2, 729, 1152]), ValueId(356): TensorShape([2, 729, 1152]), ValueId(357): TensorShape([2, 729, 1152]), ValueId(358): TensorShape([2, 729, 1152]), ValueId(359): TensorShape([2, 729, 4304]), ValueId(360): TensorShape([2, 729, 4304]), ValueId(361): TensorShape([2, 729, 4304]), ValueId(362): TensorShape([2, 729, 1152]), ValueId(363): TensorShape([2, 729, 1152]), ValueId(364): TensorShape([2, 729, 1152])} } }) }, Operation { id: OperationId(25), inputs: [ValueId(365), ValueId(366), ValueId(367)], results: [ValueId(368)], kind: LayerNorm }, Operation { id: OperationId(26), inputs: [ValueId(368), ValueId(370)], results: [ValueId(372)], kind: Gemm(GemmOptions { transpose_left: false, transpose_right: false }) }, Operation { id: OperationId(27), inputs: [ValueId(372), ValueId(371)], results: [ValueId(373)], kind: Add }, Operation { id: OperationId(28), inputs: [ValueId(369), ValueId(374)], results: [ValueId(376)], kind: Gemm(GemmOptions { transpose_left: false, transpose_right: false }) }, Operation { id: OperationId(29), inputs: [ValueId(376), ValueId(375)], results: [ValueId(377)], kind: Add }, Operation { id: OperationId(30), inputs: [ValueId(373)], results: [ValueId(378)], kind: Slice(AxisSlice { axis: 2, start: 0, length: 1152 }) }, Operation { id: OperationId(31), inputs: [ValueId(373)], results: [ValueId(379)], kind: Slice(AxisSlice { axis: 2, start: 1152, length: 1152 }) }, Operation { id: OperationId(32), inputs: [ValueId(377)], results: [ValueId(380)], kind: View(AxisFactorView { split_axis: 2, merge_axis: 0, factor: 16, reversed: false }) }, Operation { id: OperationId(33), inputs: [ValueId(378)], results: [ValueId(381)], kind: View(AxisFactorView { split_axis: 2, merge_axis: 0, factor: 16, reversed: false }) }, Operation { id: OperationId(34), inputs: [ValueId(379)], results: [ValueId(382)], kind: View(AxisFactorView { split_axis: 2, merge_axis: 0, factor: 16, reversed: false }) }, Operation { id: OperationId(35), inputs: [ValueId(380), ValueId(381), ValueId(382)], results: [ValueId(383)], kind: FlashAttention(AttentionOptions { causal: false, scale: InverseSqrtQueryWidth }) }, Operation { id: OperationId(36), inputs: [ValueId(383)], results: [ValueId(384)], kind: View(AxisFactorView { split_axis: 2, merge_axis: 0, factor: 16, reversed: true }) }, Operation { id: OperationId(37), inputs: [ValueId(384), ValueId(385)], results: [ValueId(387)], kind: Gemm(GemmOptions { transpose_left: false, transpose_right: false }) }, Operation { id: OperationId(38), inputs: [ValueId(387), ValueId(386)], results: [ValueId(388)], kind: Add }, Operation { id: OperationId(39), inputs: [ValueId(388), ValueId(389), ValueId(390)], results: [ValueId(391)], kind: LayerNorm }, Operation { id: OperationId(40), inputs: [ValueId(391), ValueId(392)], results: [ValueId(394)], kind: Gemm(GemmOptions { transpose_left: false, transpose_right: false }) }, Operation { id: OperationId(41), inputs: [ValueId(394), ValueId(393)], results: [ValueId(395)], kind: Add }, Operation { id: OperationId(42), inputs: [ValueId(395)], results: [ValueId(396)], kind: Gelu }, Operation { id: OperationId(43), inputs: [ValueId(396), ValueId(397)], results: [ValueId(399)], kind: Gemm(GemmOptions { transpose_left: false, transpose_right: false }) }, Operation { id: OperationId(44), inputs: [ValueId(399), ValueId(398)], results: [ValueId(400)], kind: Add }, Operation { id: OperationId(45), inputs: [ValueId(388), ValueId(400)], results: [ValueId(401)], kind: Add }], outputs: [ValueId(401)], values: {ValueId(0), ValueId(1), ValueId(2), ValueId(3), ValueId(4), ValueId(5), ValueId(6), ValueId(7), ValueId(8), ValueId(9), ValueId(10), ValueId(11), ValueId(12), ValueId(13), ValueId(14), ValueId(15), ValueId(16), ValueId(17), ValueId(18), ValueId(19), ValueId(20), ValueId(21), ValueId(22), ValueId(23), ValueId(24), ValueId(25), ValueId(26), ValueId(27), ValueId(28), ValueId(29), ValueId(30), ValueId(31), ValueId(32), ValueId(33), ValueId(34), ValueId(35), ValueId(36), ValueId(37), ValueId(38), ValueId(39), ValueId(40), ValueId(41), ValueId(42), ValueId(43), ValueId(44), ValueId(45), ValueId(46), ValueId(47), ValueId(48), ValueId(49), ValueId(50), ValueId(51), ValueId(52), ValueId(53), ValueId(54), ValueId(55), ValueId(56), ValueId(57), ValueId(58), ValueId(59), ValueId(60), ValueId(61), ValueId(62), ValueId(63), ValueId(64), ValueId(65), ValueId(66), ValueId(67), ValueId(68), ValueId(69), ValueId(70), ValueId(71), ValueId(72), ValueId(73), ValueId(74), ValueId(75), ValueId(76), ValueId(77), ValueId(78), ValueId(79), ValueId(80), ValueId(81), ValueId(82), ValueId(83), ValueId(84), ValueId(85), ValueId(86), ValueId(87), ValueId(88), ValueId(89), ValueId(90), ValueId(91), ValueId(92), ValueId(93), ValueId(94), ValueId(95), ValueId(96), ValueId(97), ValueId(98), ValueId(99), ValueId(100), ValueId(101), ValueId(102), ValueId(103), ValueId(104), ValueId(105), ValueId(106), ValueId(107), ValueId(108), ValueId(109), ValueId(110), ValueId(111), ValueId(112), ValueId(113), ValueId(114), ValueId(115), ValueId(116), ValueId(117), ValueId(118), ValueId(119), ValueId(120), ValueId(121), ValueId(122), ValueId(123), ValueId(124), ValueId(125), ValueId(126), ValueId(127), ValueId(128), ValueId(129), ValueId(130), ValueId(131), ValueId(132), ValueId(133), ValueId(134), ValueId(135), ValueId(136), ValueId(137), ValueId(138), ValueId(139), ValueId(140), ValueId(141), ValueId(142), ValueId(143), ValueId(144), ValueId(145), ValueId(146), ValueId(147), ValueId(148), ValueId(149), ValueId(150), ValueId(151), ValueId(152), ValueId(153), ValueId(154), ValueId(155), ValueId(156), ValueId(157), ValueId(158), ValueId(159), ValueId(160), ValueId(161), ValueId(162), ValueId(163), ValueId(164), ValueId(165), ValueId(166), ValueId(167), ValueId(168), ValueId(169), ValueId(170), ValueId(171), ValueId(172), ValueId(173), ValueId(174), ValueId(175), ValueId(176), ValueId(177), ValueId(178), ValueId(179), ValueId(180), ValueId(181), ValueId(182), ValueId(183), ValueId(184), ValueId(185), ValueId(186), ValueId(187), ValueId(188), ValueId(189), ValueId(190), ValueId(191), ValueId(192), ValueId(193), ValueId(194), ValueId(195), ValueId(196), ValueId(197), ValueId(198), ValueId(199), ValueId(200), ValueId(201), ValueId(202), ValueId(203), ValueId(204), ValueId(205), ValueId(206), ValueId(207), ValueId(208), ValueId(209), ValueId(210), ValueId(211), ValueId(212), ValueId(213), ValueId(214), ValueId(215), ValueId(216), ValueId(217), ValueId(218), ValueId(219), ValueId(220), ValueId(221), ValueId(222), ValueId(223), ValueId(224), ValueId(225), ValueId(226), ValueId(227), ValueId(228), ValueId(229), ValueId(230), ValueId(231), ValueId(232), ValueId(233), ValueId(234), ValueId(235), ValueId(236), ValueId(237), ValueId(238), ValueId(239), ValueId(240), ValueId(241), ValueId(242), ValueId(243), ValueId(244), ValueId(245), ValueId(246), ValueId(247), ValueId(248), ValueId(249), ValueId(250), ValueId(251), ValueId(252), ValueId(253), ValueId(254), ValueId(255), ValueId(256), ValueId(257), ValueId(258), ValueId(259), ValueId(260), ValueId(261), ValueId(262), ValueId(263), ValueId(264), ValueId(265), ValueId(266), ValueId(267), ValueId(268), ValueId(269), ValueId(270), ValueId(271), ValueId(272), ValueId(273), ValueId(274), ValueId(275), ValueId(276), ValueId(277), ValueId(278), ValueId(279), ValueId(280), ValueId(281), ValueId(282), ValueId(283), ValueId(284), ValueId(285), ValueId(286), ValueId(287), ValueId(288), ValueId(289), ValueId(290), ValueId(291), ValueId(292), ValueId(293), ValueId(294), ValueId(295), ValueId(296), ValueId(297), ValueId(298), ValueId(299), ValueId(300), ValueId(301), ValueId(302), ValueId(303), ValueId(304), ValueId(305), ValueId(306), ValueId(307), ValueId(308), ValueId(309), ValueId(310), ValueId(311), ValueId(312), ValueId(313), ValueId(314), ValueId(315), ValueId(316), ValueId(317), ValueId(318), ValueId(319), ValueId(320), ValueId(321), ValueId(322), ValueId(323), ValueId(324), ValueId(325), ValueId(326), ValueId(327), ValueId(328), ValueId(329), ValueId(330), ValueId(365), ValueId(366), ValueId(367), ValueId(368), ValueId(369), ValueId(370), ValueId(371), ValueId(372), ValueId(373), ValueId(374), ValueId(375), ValueId(376), ValueId(377), ValueId(378), ValueId(379), ValueId(380), ValueId(381), ValueId(382), ValueId(383), ValueId(384), ValueId(385), ValueId(386), ValueId(387), ValueId(388), ValueId(389), ValueId(390), ValueId(391), ValueId(392), ValueId(393), ValueId(394), ValueId(395), ValueId(396), ValueId(397), ValueId(398), ValueId(399), ValueId(400), ValueId(401)}, shapes: {ValueId(0): TensorShape([2, 729, 588]), ValueId(1): TensorShape([588, 1152]), ValueId(2): TensorShape([1, 1, 1152]), ValueId(3): TensorShape([2, 729, 1152]), ValueId(4): TensorShape([2, 729, 1152]), ValueId(5): TensorShape([1, 729, 1152]), ValueId(6): TensorShape([2, 729, 1152]), ValueId(7): TensorShape([1, 1, 1152]), ValueId(8): TensorShape([1, 1, 1152]), ValueId(9): TensorShape([1, 1, 1152]), ValueId(10): TensorShape([1, 1, 1152]), ValueId(11): TensorShape([1, 1, 1152]), ValueId(12): TensorShape([1, 1, 1152]), ValueId(13): TensorShape([1, 1, 1152]), ValueId(14): TensorShape([1, 1, 1152]), ValueId(15): TensorShape([1, 1, 1152]), ValueId(16): TensorShape([1, 1, 1152]), ValueId(17): TensorShape([1, 1, 1152]), ValueId(18): TensorShape([1, 1, 1152]), ValueId(19): TensorShape([1, 1, 1152]), ValueId(20): TensorShape([1, 1, 1152]), ValueId(21): TensorShape([1, 1, 1152]), ValueId(22): TensorShape([1, 1, 1152]), ValueId(23): TensorShape([1, 1, 1152]), ValueId(24): TensorShape([1, 1, 1152]), ValueId(25): TensorShape([1, 1, 1152]), ValueId(26): TensorShape([1, 1, 1152]), ValueId(27): TensorShape([1, 1, 1152]), ValueId(28): TensorShape([1, 1, 1152]), ValueId(29): TensorShape([1, 1, 1152]), ValueId(30): TensorShape([1, 1, 1152]), ValueId(31): TensorShape([1, 1, 1152]), ValueId(32): TensorShape([1, 1, 1152]), ValueId(33): TensorShape([1, 1, 1152]), ValueId(34): TensorShape([1, 1, 1152]), ValueId(35): TensorShape([1, 1, 1152]), ValueId(36): TensorShape([1, 1, 1152]), ValueId(37): TensorShape([1, 1, 1152]), ValueId(38): TensorShape([1, 1, 1152]), ValueId(39): TensorShape([1, 1, 1152]), ValueId(40): TensorShape([1, 1, 1152]), ValueId(41): TensorShape([1, 1, 1152]), ValueId(42): TensorShape([1, 1, 1152]), ValueId(43): TensorShape([1, 1, 1152]), ValueId(44): TensorShape([1, 1, 1152]), ValueId(45): TensorShape([1, 1, 1152]), ValueId(46): TensorShape([1, 1, 1152]), ValueId(47): TensorShape([1, 1, 1152]), ValueId(48): TensorShape([1, 1, 1152]), ValueId(49): TensorShape([1, 1, 1152]), ValueId(50): TensorShape([1, 1, 1152]), ValueId(51): TensorShape([1, 1, 1152]), ValueId(52): TensorShape([1, 1, 1152]), ValueId(53): TensorShape([1, 1, 1152]), ValueId(54): TensorShape([1, 1, 1152]), ValueId(55): TensorShape([1, 1, 1152]), ValueId(56): TensorShape([1, 1, 1152]), ValueId(57): TensorShape([1, 1, 1152]), ValueId(58): TensorShape([1, 1, 1152]), ValueId(59): TensorShape([1, 1, 1152]), ValueId(60): TensorShape([1, 1, 1152]), ValueId(61): TensorShape([1152, 3456]), ValueId(62): TensorShape([1152, 3456]), ValueId(63): TensorShape([1152, 3456]), ValueId(64): TensorShape([1152, 3456]), ValueId(65): TensorShape([1152, 3456]), ValueId(66): TensorShape([1152, 3456]), ValueId(67): TensorShape([1152, 3456]), ValueId(68): TensorShape([1152, 3456]), ValueId(69): TensorShape([1152, 3456]), ValueId(70): TensorShape([1152, 3456]), ValueId(71): TensorShape([1152, 3456]), ValueId(72): TensorShape([1152, 3456]), ValueId(73): TensorShape([1152, 3456]), ValueId(74): TensorShape([1152, 3456]), ValueId(75): TensorShape([1152, 3456]), ValueId(76): TensorShape([1152, 3456]), ValueId(77): TensorShape([1152, 3456]), ValueId(78): TensorShape([1152, 3456]), ValueId(79): TensorShape([1152, 3456]), ValueId(80): TensorShape([1152, 3456]), ValueId(81): TensorShape([1152, 3456]), ValueId(82): TensorShape([1152, 3456]), ValueId(83): TensorShape([1152, 3456]), ValueId(84): TensorShape([1152, 3456]), ValueId(85): TensorShape([1152, 3456]), ValueId(86): TensorShape([1152, 3456]), ValueId(87): TensorShape([1152, 3456]), ValueId(88): TensorShape([1, 1, 3456]), ValueId(89): TensorShape([1, 1, 3456]), ValueId(90): TensorShape([1, 1, 3456]), ValueId(91): TensorShape([1, 1, 3456]), ValueId(92): TensorShape([1, 1, 3456]), ValueId(93): TensorShape([1, 1, 3456]), ValueId(94): TensorShape([1, 1, 3456]), ValueId(95): TensorShape([1, 1, 3456]), ValueId(96): TensorShape([1, 1, 3456]), ValueId(97): TensorShape([1, 1, 3456]), ValueId(98): TensorShape([1, 1, 3456]), ValueId(99): TensorShape([1, 1, 3456]), ValueId(100): TensorShape([1, 1, 3456]), ValueId(101): TensorShape([1, 1, 3456]), ValueId(102): TensorShape([1, 1, 3456]), ValueId(103): TensorShape([1, 1, 3456]), ValueId(104): TensorShape([1, 1, 3456]), ValueId(105): TensorShape([1, 1, 3456]), ValueId(106): TensorShape([1, 1, 3456]), ValueId(107): TensorShape([1, 1, 3456]), ValueId(108): TensorShape([1, 1, 3456]), ValueId(109): TensorShape([1, 1, 3456]), ValueId(110): TensorShape([1, 1, 3456]), ValueId(111): TensorShape([1, 1, 3456]), ValueId(112): TensorShape([1, 1, 3456]), ValueId(113): TensorShape([1, 1, 3456]), ValueId(114): TensorShape([1, 1, 3456]), ValueId(115): TensorShape([1152, 1152]), ValueId(116): TensorShape([1152, 1152]), ValueId(117): TensorShape([1152, 1152]), ValueId(118): TensorShape([1152, 1152]), ValueId(119): TensorShape([1152, 1152]), ValueId(120): TensorShape([1152, 1152]), ValueId(121): TensorShape([1152, 1152]), ValueId(122): TensorShape([1152, 1152]), ValueId(123): TensorShape([1152, 1152]), ValueId(124): TensorShape([1152, 1152]), ValueId(125): TensorShape([1152, 1152]), ValueId(126): TensorShape([1152, 1152]), ValueId(127): TensorShape([1152, 1152]), ValueId(128): TensorShape([1152, 1152]), ValueId(129): TensorShape([1152, 1152]), ValueId(130): TensorShape([1152, 1152]), ValueId(131): TensorShape([1152, 1152]), ValueId(132): TensorShape([1152, 1152]), ValueId(133): TensorShape([1152, 1152]), ValueId(134): TensorShape([1152, 1152]), ValueId(135): TensorShape([1152, 1152]), ValueId(136): TensorShape([1152, 1152]), ValueId(137): TensorShape([1152, 1152]), ValueId(138): TensorShape([1152, 1152]), ValueId(139): TensorShape([1152, 1152]), ValueId(140): TensorShape([1152, 1152]), ValueId(141): TensorShape([1152, 1152]), ValueId(142): TensorShape([1, 1, 1152]), ValueId(143): TensorShape([1, 1, 1152]), ValueId(144): TensorShape([1, 1, 1152]), ValueId(145): TensorShape([1, 1, 1152]), ValueId(146): TensorShape([1, 1, 1152]), ValueId(147): TensorShape([1, 1, 1152]), ValueId(148): TensorShape([1, 1, 1152]), ValueId(149): TensorShape([1, 1, 1152]), ValueId(150): TensorShape([1, 1, 1152]), ValueId(151): TensorShape([1, 1, 1152]), ValueId(152): TensorShape([1, 1, 1152]), ValueId(153): TensorShape([1, 1, 1152]), ValueId(154): TensorShape([1, 1, 1152]), ValueId(155): TensorShape([1, 1, 1152]), ValueId(156): TensorShape([1, 1, 1152]), ValueId(157): TensorShape([1, 1, 1152]), ValueId(158): TensorShape([1, 1, 1152]), ValueId(159): TensorShape([1, 1, 1152]), ValueId(160): TensorShape([1, 1, 1152]), ValueId(161): TensorShape([1, 1, 1152]), ValueId(162): TensorShape([1, 1, 1152]), ValueId(163): TensorShape([1, 1, 1152]), ValueId(164): TensorShape([1, 1, 1152]), ValueId(165): TensorShape([1, 1, 1152]), ValueId(166): TensorShape([1, 1, 1152]), ValueId(167): TensorShape([1, 1, 1152]), ValueId(168): TensorShape([1, 1, 1152]), ValueId(169): TensorShape([1, 1, 1152]), ValueId(170): TensorShape([1, 1, 1152]), ValueId(171): TensorShape([1, 1, 1152]), ValueId(172): TensorShape([1, 1, 1152]), ValueId(173): TensorShape([1, 1, 1152]), ValueId(174): TensorShape([1, 1, 1152]), ValueId(175): TensorShape([1, 1, 1152]), ValueId(176): TensorShape([1, 1, 1152]), ValueId(177): TensorShape([1, 1, 1152]), ValueId(178): TensorShape([1, 1, 1152]), ValueId(179): TensorShape([1, 1, 1152]), ValueId(180): TensorShape([1, 1, 1152]), ValueId(181): TensorShape([1, 1, 1152]), ValueId(182): TensorShape([1, 1, 1152]), ValueId(183): TensorShape([1, 1, 1152]), ValueId(184): TensorShape([1, 1, 1152]), ValueId(185): TensorShape([1, 1, 1152]), ValueId(186): TensorShape([1, 1, 1152]), ValueId(187): TensorShape([1, 1, 1152]), ValueId(188): TensorShape([1, 1, 1152]), ValueId(189): TensorShape([1, 1, 1152]), ValueId(190): TensorShape([1, 1, 1152]), ValueId(191): TensorShape([1, 1, 1152]), ValueId(192): TensorShape([1, 1, 1152]), ValueId(193): TensorShape([1, 1, 1152]), ValueId(194): TensorShape([1, 1, 1152]), ValueId(195): TensorShape([1, 1, 1152]), ValueId(196): TensorShape([1, 1, 1152]), ValueId(197): TensorShape([1, 1, 1152]), ValueId(198): TensorShape([1, 1, 1152]), ValueId(199): TensorShape([1, 1, 1152]), ValueId(200): TensorShape([1, 1, 1152]), ValueId(201): TensorShape([1, 1, 1152]), ValueId(202): TensorShape([1, 1, 1152]), ValueId(203): TensorShape([1, 1, 1152]), ValueId(204): TensorShape([1, 1, 1152]), ValueId(205): TensorShape([1, 1, 1152]), ValueId(206): TensorShape([1, 1, 1152]), ValueId(207): TensorShape([1, 1, 1152]), ValueId(208): TensorShape([1, 1, 1152]), ValueId(209): TensorShape([1, 1, 1152]), ValueId(210): TensorShape([1, 1, 1152]), ValueId(211): TensorShape([1, 1, 1152]), ValueId(212): TensorShape([1, 1, 1152]), ValueId(213): TensorShape([1, 1, 1152]), ValueId(214): TensorShape([1, 1, 1152]), ValueId(215): TensorShape([1, 1, 1152]), ValueId(216): TensorShape([1, 1, 1152]), ValueId(217): TensorShape([1, 1, 1152]), ValueId(218): TensorShape([1, 1, 1152]), ValueId(219): TensorShape([1, 1, 1152]), ValueId(220): TensorShape([1, 1, 1152]), ValueId(221): TensorShape([1, 1, 1152]), ValueId(222): TensorShape([1, 1, 1152]), ValueId(223): TensorShape([1152, 4304]), ValueId(224): TensorShape([1152, 4304]), ValueId(225): TensorShape([1152, 4304]), ValueId(226): TensorShape([1152, 4304]), ValueId(227): TensorShape([1152, 4304]), ValueId(228): TensorShape([1152, 4304]), ValueId(229): TensorShape([1152, 4304]), ValueId(230): TensorShape([1152, 4304]), ValueId(231): TensorShape([1152, 4304]), ValueId(232): TensorShape([1152, 4304]), ValueId(233): TensorShape([1152, 4304]), ValueId(234): TensorShape([1152, 4304]), ValueId(235): TensorShape([1152, 4304]), ValueId(236): TensorShape([1152, 4304]), ValueId(237): TensorShape([1152, 4304]), ValueId(238): TensorShape([1152, 4304]), ValueId(239): TensorShape([1152, 4304]), ValueId(240): TensorShape([1152, 4304]), ValueId(241): TensorShape([1152, 4304]), ValueId(242): TensorShape([1152, 4304]), ValueId(243): TensorShape([1152, 4304]), ValueId(244): TensorShape([1152, 4304]), ValueId(245): TensorShape([1152, 4304]), ValueId(246): TensorShape([1152, 4304]), ValueId(247): TensorShape([1152, 4304]), ValueId(248): TensorShape([1152, 4304]), ValueId(249): TensorShape([1152, 4304]), ValueId(250): TensorShape([1, 1, 4304]), ValueId(251): TensorShape([1, 1, 4304]), ValueId(252): TensorShape([1, 1, 4304]), ValueId(253): TensorShape([1, 1, 4304]), ValueId(254): TensorShape([1, 1, 4304]), ValueId(255): TensorShape([1, 1, 4304]), ValueId(256): TensorShape([1, 1, 4304]), ValueId(257): TensorShape([1, 1, 4304]), ValueId(258): TensorShape([1, 1, 4304]), ValueId(259): TensorShape([1, 1, 4304]), ValueId(260): TensorShape([1, 1, 4304]), ValueId(261): TensorShape([1, 1, 4304]), ValueId(262): TensorShape([1, 1, 4304]), ValueId(263): TensorShape([1, 1, 4304]), ValueId(264): TensorShape([1, 1, 4304]), ValueId(265): TensorShape([1, 1, 4304]), ValueId(266): TensorShape([1, 1, 4304]), ValueId(267): TensorShape([1, 1, 4304]), ValueId(268): TensorShape([1, 1, 4304]), ValueId(269): TensorShape([1, 1, 4304]), ValueId(270): TensorShape([1, 1, 4304]), ValueId(271): TensorShape([1, 1, 4304]), ValueId(272): TensorShape([1, 1, 4304]), ValueId(273): TensorShape([1, 1, 4304]), ValueId(274): TensorShape([1, 1, 4304]), ValueId(275): TensorShape([1, 1, 4304]), ValueId(276): TensorShape([1, 1, 4304]), ValueId(277): TensorShape([4304, 1152]), ValueId(278): TensorShape([4304, 1152]), ValueId(279): TensorShape([4304, 1152]), ValueId(280): TensorShape([4304, 1152]), ValueId(281): TensorShape([4304, 1152]), ValueId(282): TensorShape([4304, 1152]), ValueId(283): TensorShape([4304, 1152]), ValueId(284): TensorShape([4304, 1152]), ValueId(285): TensorShape([4304, 1152]), ValueId(286): TensorShape([4304, 1152]), ValueId(287): TensorShape([4304, 1152]), ValueId(288): TensorShape([4304, 1152]), ValueId(289): TensorShape([4304, 1152]), ValueId(290): TensorShape([4304, 1152]), ValueId(291): TensorShape([4304, 1152]), ValueId(292): TensorShape([4304, 1152]), ValueId(293): TensorShape([4304, 1152]), ValueId(294): TensorShape([4304, 1152]), ValueId(295): TensorShape([4304, 1152]), ValueId(296): TensorShape([4304, 1152]), ValueId(297): TensorShape([4304, 1152]), ValueId(298): TensorShape([4304, 1152]), ValueId(299): TensorShape([4304, 1152]), ValueId(300): TensorShape([4304, 1152]), ValueId(301): TensorShape([4304, 1152]), ValueId(302): TensorShape([4304, 1152]), ValueId(303): TensorShape([4304, 1152]), ValueId(304): TensorShape([1, 1, 1152]), ValueId(305): TensorShape([1, 1, 1152]), ValueId(306): TensorShape([1, 1, 1152]), ValueId(307): TensorShape([1, 1, 1152]), ValueId(308): TensorShape([1, 1, 1152]), ValueId(309): TensorShape([1, 1, 1152]), ValueId(310): TensorShape([1, 1, 1152]), ValueId(311): TensorShape([1, 1, 1152]), ValueId(312): TensorShape([1, 1, 1152]), ValueId(313): TensorShape([1, 1, 1152]), ValueId(314): TensorShape([1, 1, 1152]), ValueId(315): TensorShape([1, 1, 1152]), ValueId(316): TensorShape([1, 1, 1152]), ValueId(317): TensorShape([1, 1, 1152]), ValueId(318): TensorShape([1, 1, 1152]), ValueId(319): TensorShape([1, 1, 1152]), ValueId(320): TensorShape([1, 1, 1152]), ValueId(321): TensorShape([1, 1, 1152]), ValueId(322): TensorShape([1, 1, 1152]), ValueId(323): TensorShape([1, 1, 1152]), ValueId(324): TensorShape([1, 1, 1152]), ValueId(325): TensorShape([1, 1, 1152]), ValueId(326): TensorShape([1, 1, 1152]), ValueId(327): TensorShape([1, 1, 1152]), ValueId(328): TensorShape([1, 1, 1152]), ValueId(329): TensorShape([1, 1, 1152]), ValueId(330): TensorShape([1, 1, 1152]), ValueId(365): TensorShape([2, 729, 1152]), ValueId(366): TensorShape([1, 1, 1152]), ValueId(367): TensorShape([1, 1, 1152]), ValueId(368): TensorShape([2, 729, 1152]), ValueId(369): TensorShape([2, 1, 1152]), ValueId(370): TensorShape([1152, 2304]), ValueId(371): TensorShape([1, 1, 2304]), ValueId(372): TensorShape([2, 729, 2304]), ValueId(373): TensorShape([2, 729, 2304]), ValueId(374): TensorShape([1152, 1152]), ValueId(375): TensorShape([1, 1, 1152]), ValueId(376): TensorShape([2, 1, 1152]), ValueId(377): TensorShape([2, 1, 1152]), ValueId(378): TensorShape([2, 729, 1152]), ValueId(379): TensorShape([2, 729, 1152]), ValueId(380): TensorShape([32, 1, 72]), ValueId(381): TensorShape([32, 729, 72]), ValueId(382): TensorShape([32, 729, 72]), ValueId(383): TensorShape([32, 1, 72]), ValueId(384): TensorShape([2, 1, 1152]), ValueId(385): TensorShape([1152, 1152]), ValueId(386): TensorShape([1, 1, 1152]), ValueId(387): TensorShape([2, 1, 1152]), ValueId(388): TensorShape([2, 1, 1152]), ValueId(389): TensorShape([1, 1, 1152]), ValueId(390): TensorShape([1, 1, 1152]), ValueId(391): TensorShape([2, 1, 1152]), ValueId(392): TensorShape([1152, 4304]), ValueId(393): TensorShape([1, 1, 4304]), ValueId(394): TensorShape([2, 1, 4304]), ValueId(395): TensorShape([2, 1, 4304]), ValueId(396): TensorShape([2, 1, 4304]), ValueId(397): TensorShape([4304, 1152]), ValueId(398): TensorShape([1, 1, 1152]), ValueId(399): TensorShape([2, 1, 1152]), ValueId(400): TensorShape([2, 1, 1152]), ValueId(401): TensorShape([2, 1, 1152])}, next_operation: 46, next_value: 402, next_sequence: 12 }\nPipelineConfig { tile_count: 1472, optimization_steps: 0, load_search_state: None, save_search_state: None, capacity_baseline: true, exchange_stream_words: Some(1024), inputs: {}, automatic_inputs: {ValueId(0): F16, ValueId(1): F8F143 { scale_exponent: -4 }, ValueId(2): F16, ValueId(5): F16, ValueId(7): F16, ValueId(8): F16, ValueId(9): F16, ValueId(10): F16, ValueId(11): F16, ValueId(12): F16, ValueId(13): F16, ValueId(14): F16, ValueId(15): F16, ValueId(16): F16, ValueId(17): F16, ValueId(18): F16, ValueId(19): F16, ValueId(20): F16, ValueId(21): F16, ValueId(22): F16, ValueId(23): F16, ValueId(24): F16, ValueId(25): F16, ValueId(26): F16, ValueId(27): F16, ValueId(28): F16, ValueId(29): F16, ValueId(30): F16, ValueId(31): F16, ValueId(32): F16, ValueId(33): F16, ValueId(34): F16, ValueId(35): F16, ValueId(36): F16, ValueId(37): F16, ValueId(38): F16, ValueId(39): F16, ValueId(40): F16, ValueId(41): F16, ValueId(42): F16, ValueId(43): F16, ValueId(44): F16, ValueId(45): F16, ValueId(46): F16, ValueId(47): F16, ValueId(48): F16, ValueId(49): F16, ValueId(50): F16, ValueId(51): F16, ValueId(52): F16, ValueId(53): F16, ValueId(54): F16, ValueId(55): F16, ValueId(56): F16, ValueId(57): F16, ValueId(58): F16, ValueId(59): F16, ValueId(60): F16, ValueId(61): F8F143 { scale_exponent: -4 }, ValueId(62): F8F143 { scale_exponent: -4 }, ValueId(63): F8F143 { scale_exponent: -4 }, ValueId(64): F8F143 { scale_exponent: -4 }, ValueId(65): F8F143 { scale_exponent: -4 }, ValueId(66): F8F143 { scale_exponent: -4 }, ValueId(67): F8F143 { scale_exponent: -4 }, ValueId(68): F8F143 { scale_exponent: -4 }, ValueId(69): F8F143 { scale_exponent: -4 }, ValueId(70): F8F143 { scale_exponent: -4 }, ValueId(71): F8F143 { scale_exponent: -4 }, ValueId(72): F8F143 { scale_exponent: -4 }, ValueId(73): F8F143 { scale_exponent: -4 }, ValueId(74): F8F143 { scale_exponent: -4 }, ValueId(75): F8F143 { scale_exponent: -4 }, ValueId(76): F8F143 { scale_exponent: -4 }, ValueId(77): F8F143 { scale_exponent: -4 }, ValueId(78): F8F143 { scale_exponent: -4 }, ValueId(79): F8F143 { scale_exponent: -4 }, ValueId(80): F8F143 { scale_exponent: -4 }, ValueId(81): F8F143 { scale_exponent: -4 }, ValueId(82): F8F143 { scale_exponent: -4 }, ValueId(83): F8F143 { scale_exponent: -4 }, ValueId(84): F8F143 { scale_exponent: -4 }, ValueId(85): F8F143 { scale_exponent: -4 }, ValueId(86): F8F143 { scale_exponent: -4 }, ValueId(87): F8F143 { scale_exponent: -4 }, ValueId(88): F16, ValueId(89): F16, ValueId(90): F16, ValueId(91): F16, ValueId(92): F16, ValueId(93): F16, ValueId(94): F16, ValueId(95): F16, ValueId(96): F16, ValueId(97): F16, ValueId(98): F16, ValueId(99): F16, ValueId(100): F16, ValueId(101): F16, ValueId(102): F16, ValueId(103): F16, ValueId(104): F16, ValueId(105): F16, ValueId(106): F16, ValueId(107): F16, ValueId(108): F16, ValueId(109): F16, ValueId(110): F16, ValueId(111): F16, ValueId(112): F16, ValueId(113): F16, ValueId(114): F16, ValueId(115): F8F143 { scale_exponent: -4 }, ValueId(116): F8F143 { scale_exponent: -4 }, ValueId(117): F8F143 { scale_exponent: -4 }, ValueId(118): F8F143 { scale_exponent: -4 }, ValueId(119): F8F143 { scale_exponent: -4 }, ValueId(120): F8F143 { scale_exponent: -4 }, ValueId(121): F8F143 { scale_exponent: -4 }, ValueId(122): F8F143 { scale_exponent: -4 }, ValueId(123): F8F143 { scale_exponent: -4 }, ValueId(124): F8F143 { scale_exponent: -4 }, ValueId(125): F8F143 { scale_exponent: -4 }, ValueId(126): F8F143 { scale_exponent: -4 }, ValueId(127): F8F143 { scale_exponent: -4 }, ValueId(128): F8F143 { scale_exponent: -4 }, ValueId(129): F8F143 { scale_exponent: -4 }, ValueId(130): F8F143 { scale_exponent: -4 }, ValueId(131): F8F143 { scale_exponent: -4 }, ValueId(132): F8F143 { scale_exponent: -4 }, ValueId(133): F8F143 { scale_exponent: -4 }, ValueId(134): F8F143 { scale_exponent: -4 }, ValueId(135): F8F143 { scale_exponent: -4 }, ValueId(136): F8F143 { scale_exponent: -4 }, ValueId(137): F8F143 { scale_exponent: -4 }, ValueId(138): F8F143 { scale_exponent: -4 }, ValueId(139): F8F143 { scale_exponent: -4 }, ValueId(140): F8F143 { scale_exponent: -4 }, ValueId(141): F8F143 { scale_exponent: -4 }, ValueId(142): F16, ValueId(143): F16, ValueId(144): F16, ValueId(145): F16, ValueId(146): F16, ValueId(147): F16, ValueId(148): F16, ValueId(149): F16, ValueId(150): F16, ValueId(151): F16, ValueId(152): F16, ValueId(153): F16, ValueId(154): F16, ValueId(155): F16, ValueId(156): F16, ValueId(157): F16, ValueId(158): F16, ValueId(159): F16, ValueId(160): F16, ValueId(161): F16, ValueId(162): F16, ValueId(163): F16, ValueId(164): F16, ValueId(165): F16, ValueId(166): F16, ValueId(167): F16, ValueId(168): F16, ValueId(169): F16, ValueId(170): F16, ValueId(171): F16, ValueId(172): F16, ValueId(173): F16, ValueId(174): F16, ValueId(175): F16, ValueId(176): F16, ValueId(177): F16, ValueId(178): F16, ValueId(179): F16, ValueId(180): F16, ValueId(181): F16, ValueId(182): F16, ValueId(183): F16, ValueId(184): F16, ValueId(185): F16, ValueId(186): F16, ValueId(187): F16, ValueId(188): F16, ValueId(189): F16, ValueId(190): F16, ValueId(191): F16, ValueId(192): F16, ValueId(193): F16, ValueId(194): F16, ValueId(195): F16, ValueId(196): F16, ValueId(197): F16, ValueId(198): F16, ValueId(199): F16, ValueId(200): F16, ValueId(201): F16, ValueId(202): F16, ValueId(203): F16, ValueId(204): F16, ValueId(205): F16, ValueId(206): F16, ValueId(207): F16, ValueId(208): F16, ValueId(209): F16, ValueId(210): F16, ValueId(211): F16, ValueId(212): F16, ValueId(213): F16, ValueId(214): F16, ValueId(215): F16, ValueId(216): F16, ValueId(217): F16, ValueId(218): F16, ValueId(219): F16, ValueId(220): F16, ValueId(221): F16, ValueId(222): F16, ValueId(223): F8F143 { scale_exponent: -4 }, ValueId(224): F8F143 { scale_exponent: -4 }, ValueId(225): F8F143 { scale_exponent: -4 }, ValueId(226): F8F143 { scale_exponent: -4 }, ValueId(227): F8F143 { scale_exponent: -4 }, ValueId(228): F8F143 { scale_exponent: -4 }, ValueId(229): F8F143 { scale_exponent: -4 }, ValueId(230): F8F143 { scale_exponent: -4 }, ValueId(231): F8F143 { scale_exponent: -4 }, ValueId(232): F8F143 { scale_exponent: -4 }, ValueId(233): F8F143 { scale_exponent: -4 }, ValueId(234): F8F143 { scale_exponent: -4 }, ValueId(235): F8F143 { scale_exponent: -4 }, ValueId(236): F8F143 { scale_exponent: -4 }, ValueId(237): F8F143 { scale_exponent: -4 }, ValueId(238): F8F143 { scale_exponent: -4 }, ValueId(239): F8F143 { scale_exponent: -4 }, ValueId(240): F8F143 { scale_exponent: -4 }, ValueId(241): F8F143 { scale_exponent: -4 }, ValueId(242): F8F143 { scale_exponent: -4 }, ValueId(243): F8F143 { scale_exponent: -4 }, ValueId(244): F8F143 { scale_exponent: -4 }, ValueId(245): F8F143 { scale_exponent: -4 }, ValueId(246): F8F143 { scale_exponent: -4 }, ValueId(247): F8F143 { scale_exponent: -4 }, ValueId(248): F8F143 { scale_exponent: -4 }, ValueId(249): F8F143 { scale_exponent: -4 }, ValueId(250): F16, ValueId(251): F16, ValueId(252): F16, ValueId(253): F16, ValueId(254): F16, ValueId(255): F16, ValueId(256): F16, ValueId(257): F16, ValueId(258): F16, ValueId(259): F16, ValueId(260): F16, ValueId(261): F16, ValueId(262): F16, ValueId(263): F16, ValueId(264): F16, ValueId(265): F16, ValueId(266): F16, ValueId(267): F16, ValueId(268): F16, ValueId(269): F16, ValueId(270): F16, ValueId(271): F16, ValueId(272): F16, ValueId(273): F16, ValueId(274): F16, ValueId(275): F16, ValueId(276): F16, ValueId(277): F8F143 { scale_exponent: -4 }, ValueId(278): F8F143 { scale_exponent: -4 }, ValueId(279): F8F143 { scale_exponent: -4 }, ValueId(280): F8F143 { scale_exponent: -4 }, ValueId(281): F8F143 { scale_exponent: -4 }, ValueId(282): F8F143 { scale_exponent: -4 }, ValueId(283): F8F143 { scale_exponent: -4 }, ValueId(284): F8F143 { scale_exponent: -4 }, ValueId(285): F8F143 { scale_exponent: -4 }, ValueId(286): F8F143 { scale_exponent: -4 }, ValueId(287): F8F143 { scale_exponent: -4 }, ValueId(288): F8F143 { scale_exponent: -4 }, ValueId(289): F8F143 { scale_exponent: -4 }, ValueId(290): F8F143 { scale_exponent: -4 }, ValueId(291): F8F143 { scale_exponent: -4 }, ValueId(292): F8F143 { scale_exponent: -4 }, ValueId(293): F8F143 { scale_exponent: -4 }, ValueId(294): F8F143 { scale_exponent: -4 }, ValueId(295): F8F143 { scale_exponent: -4 }, ValueId(296): F8F143 { scale_exponent: -4 }, ValueId(297): F8F143 { scale_exponent: -4 }, ValueId(298): F8F143 { scale_exponent: -4 }, ValueId(299): F8F143 { scale_exponent: -4 }, ValueId(300): F8F143 { scale_exponent: -4 }, ValueId(301): F8F143 { scale_exponent: -4 }, ValueId(302): F8F143 { scale_exponent: -4 }, ValueId(303): F8F143 { scale_exponent: -4 }, ValueId(304): F16, ValueId(305): F16, ValueId(306): F16, ValueId(307): F16, ValueId(308): F16, ValueId(309): F16, ValueId(310): F16, ValueId(311): F16, ValueId(312): F16, ValueId(313): F16, ValueId(314): F16, ValueId(315): F16, ValueId(316): F16, ValueId(317): F16, ValueId(318): F16, ValueId(319): F16, ValueId(320): F16, ValueId(321): F16, ValueId(322): F16, ValueId(323): F16, ValueId(324): F16, ValueId(325): F16, ValueId(326): F16, ValueId(327): F16, ValueId(328): F16, ValueId(329): F16, ValueId(330): F16, ValueId(366): F16, ValueId(367): F16, ValueId(369): F16, ValueId(370): F8F143 { scale_exponent: -4 }, ValueId(371): F16, ValueId(374): F8F143 { scale_exponent: -4 }, ValueId(375): F16, ValueId(385): F8F143 { scale_exponent: -4 }, ValueId(386): F16, ValueId(389): F16, ValueId(390): F16, ValueId(392): F8F143 { scale_exponent: -4 }, ValueId(393): F16, ValueId(397): F8F143 { scale_exponent: -4 }, ValueId(398): F16}, operator_candidates: [Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Add, dispatch: Pointwise { kernel: Add }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1472, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1472, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1472, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1472, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1472, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1472, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0, 1]) }, deferred_output: None }, format_policy: PreserveInputLayout(0) }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Add, dispatch: Pointwise { kernel: Add }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1472, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1472, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1472, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1472, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1472, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1472, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0, 1]) }, deferred_output: None }, format_policy: PreserveInputLayout(1) }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Add, dispatch: Pointwise { kernel: Add }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1472, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1472, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1472, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1472, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1472, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1472, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0, 1]) }, deferred_output: None }, format_policy: RowMajorGrid }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: LayerNorm, dispatch: Pointwise { kernel: LayerNorm }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1472, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1472, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1472, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1472, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1472, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1472, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1472, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1472, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: Fresh }, deferred_output: None }, format_policy: RowMajorRows }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: LayerNorm, dispatch: Pointwise { kernel: LayerNorm }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1472, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1472, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1472, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1472, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1472, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1472, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1472, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1472, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: Fresh }, deferred_output: None }, format_policy: Concrete }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Gelu, dispatch: Pointwise { kernel: Gelu }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: Amp(Left), tiling: TensorTiling { tile_count: 1472, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1472, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }, AxisTiling { axis: FromEnd(1), partitions: 1, padding_groups: 1, block_size: 64, padding_multiple: 64, shard_padding_multiple: 1, padding: Zero, tile_stride: None }] }, memory_class: Ipu21Interleaved } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: Amp(Left), tiling: TensorTiling { tile_count: 1472, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1472, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }, AxisTiling { axis: FromEnd(1), partitions: 1, padding_groups: 1, block_size: 64, padding_multiple: 64, shard_padding_multiple: 1, padding: Zero, tile_stride: None }] }, memory_class: Ipu21Interleaved } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0]) }, deferred_output: None }, format_policy: PreserveInputLayout(0) }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Gelu, dispatch: Pointwise { kernel: Gelu }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1472, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1472, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1472, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1472, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0]) }, deferred_output: None }, format_policy: PreserveInputLayout(0) }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Add, dispatch: Pointwise { kernel: Add }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1472, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1472, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1472, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1472, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1472, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1472, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0, 1]) }, deferred_output: None }, format_policy: Concrete }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: FlashAttention { options: AttentionOptions { causal: false, scale: InverseSqrtQueryWidth }, accumulate: F32 }, dispatch: Pointwise { kernel: FlashAttention { options: AttentionOptions { causal: false, scale: InverseSqrtQueryWidth }, accumulate: F32 } }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1472, replicas: 1, axes: [AxisTiling { axis: FromEnd(3), partitions: 1472, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1472, replicas: 1, axes: [AxisTiling { axis: FromEnd(3), partitions: 1472, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1472, replicas: 1, axes: [AxisTiling { axis: FromEnd(3), partitions: 1472, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1472, replicas: 1, axes: [AxisTiling { axis: FromEnd(3), partitions: 1472, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: Fresh }, deferred_output: None }, format_policy: Concrete }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Add, dispatch: Pointwise { kernel: Add }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1024, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1024, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1024, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1024, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1024, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1024, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0, 1]) }, deferred_output: None }, format_policy: PreserveInputLayout(0) }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Add, dispatch: Pointwise { kernel: Add }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1024, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1024, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1024, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1024, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1024, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1024, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0, 1]) }, deferred_output: None }, format_policy: PreserveInputLayout(1) }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Add, dispatch: Pointwise { kernel: Add }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1024, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1024, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1024, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1024, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1024, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1024, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0, 1]) }, deferred_output: None }, format_policy: RowMajorGrid }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: LayerNorm, dispatch: Pointwise { kernel: LayerNorm }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1024, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1024, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1024, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1024, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1024, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1024, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1024, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1024, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: Fresh }, deferred_output: None }, format_policy: RowMajorRows }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: LayerNorm, dispatch: Pointwise { kernel: LayerNorm }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1024, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1024, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1024, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1024, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1024, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1024, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1024, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1024, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: Fresh }, deferred_output: None }, format_policy: Concrete }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Gelu, dispatch: Pointwise { kernel: Gelu }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: Amp(Left), tiling: TensorTiling { tile_count: 1024, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1024, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }, AxisTiling { axis: FromEnd(1), partitions: 1, padding_groups: 1, block_size: 64, padding_multiple: 64, shard_padding_multiple: 1, padding: Zero, tile_stride: None }] }, memory_class: Ipu21Interleaved } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: Amp(Left), tiling: TensorTiling { tile_count: 1024, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1024, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }, AxisTiling { axis: FromEnd(1), partitions: 1, padding_groups: 1, block_size: 64, padding_multiple: 64, shard_padding_multiple: 1, padding: Zero, tile_stride: None }] }, memory_class: Ipu21Interleaved } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0]) }, deferred_output: None }, format_policy: PreserveInputLayout(0) }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Gelu, dispatch: Pointwise { kernel: Gelu }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1024, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1024, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1024, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1024, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0]) }, deferred_output: None }, format_policy: PreserveInputLayout(0) }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Add, dispatch: Pointwise { kernel: Add }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1024, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1024, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1024, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1024, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1024, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1024, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0, 1]) }, deferred_output: None }, format_policy: Concrete }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: FlashAttention { options: AttentionOptions { causal: false, scale: InverseSqrtQueryWidth }, accumulate: F32 }, dispatch: Pointwise { kernel: FlashAttention { options: AttentionOptions { causal: false, scale: InverseSqrtQueryWidth }, accumulate: F32 } }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1024, replicas: 1, axes: [AxisTiling { axis: FromEnd(3), partitions: 1024, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1024, replicas: 1, axes: [AxisTiling { axis: FromEnd(3), partitions: 1024, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1024, replicas: 1, axes: [AxisTiling { axis: FromEnd(3), partitions: 1024, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1024, replicas: 1, axes: [AxisTiling { axis: FromEnd(3), partitions: 1024, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: Fresh }, deferred_output: None }, format_policy: Concrete }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Add, dispatch: Pointwise { kernel: Add }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 512, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 512, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 512, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 512, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 512, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 512, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0, 1]) }, deferred_output: None }, format_policy: PreserveInputLayout(0) }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Add, dispatch: Pointwise { kernel: Add }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 512, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 512, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 512, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 512, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 512, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 512, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0, 1]) }, deferred_output: None }, format_policy: PreserveInputLayout(1) }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Add, dispatch: Pointwise { kernel: Add }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 512, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 512, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 512, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 512, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 512, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 512, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0, 1]) }, deferred_output: None }, format_policy: RowMajorGrid }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: LayerNorm, dispatch: Pointwise { kernel: LayerNorm }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 512, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 512, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 512, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 512, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 512, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 512, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 512, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 512, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: Fresh }, deferred_output: None }, format_policy: RowMajorRows }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: LayerNorm, dispatch: Pointwise { kernel: LayerNorm }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 512, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 512, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 512, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 512, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 512, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 512, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 512, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 512, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: Fresh }, deferred_output: None }, format_policy: Concrete }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Gelu, dispatch: Pointwise { kernel: Gelu }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: Amp(Left), tiling: TensorTiling { tile_count: 512, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 512, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }, AxisTiling { axis: FromEnd(1), partitions: 1, padding_groups: 1, block_size: 64, padding_multiple: 64, shard_padding_multiple: 1, padding: Zero, tile_stride: None }] }, memory_class: Ipu21Interleaved } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: Amp(Left), tiling: TensorTiling { tile_count: 512, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 512, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }, AxisTiling { axis: FromEnd(1), partitions: 1, padding_groups: 1, block_size: 64, padding_multiple: 64, shard_padding_multiple: 1, padding: Zero, tile_stride: None }] }, memory_class: Ipu21Interleaved } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0]) }, deferred_output: None }, format_policy: PreserveInputLayout(0) }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Gelu, dispatch: Pointwise { kernel: Gelu }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 512, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 512, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 512, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 512, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0]) }, deferred_output: None }, format_policy: PreserveInputLayout(0) }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Add, dispatch: Pointwise { kernel: Add }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 512, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 512, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 512, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 512, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 512, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 512, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0, 1]) }, deferred_output: None }, format_policy: Concrete }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: FlashAttention { options: AttentionOptions { causal: false, scale: InverseSqrtQueryWidth }, accumulate: F32 }, dispatch: Pointwise { kernel: FlashAttention { options: AttentionOptions { causal: false, scale: InverseSqrtQueryWidth }, accumulate: F32 } }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 512, replicas: 1, axes: [AxisTiling { axis: FromEnd(3), partitions: 512, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 512, replicas: 1, axes: [AxisTiling { axis: FromEnd(3), partitions: 512, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 512, replicas: 1, axes: [AxisTiling { axis: FromEnd(3), partitions: 512, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 512, replicas: 1, axes: [AxisTiling { axis: FromEnd(3), partitions: 512, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: Fresh }, deferred_output: None }, format_policy: Concrete }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Add, dispatch: Pointwise { kernel: Add }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 256, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 256, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 256, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 256, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 256, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 256, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0, 1]) }, deferred_output: None }, format_policy: PreserveInputLayout(0) }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Add, dispatch: Pointwise { kernel: Add }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 256, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 256, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 256, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 256, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 256, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 256, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0, 1]) }, deferred_output: None }, format_policy: PreserveInputLayout(1) }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Add, dispatch: Pointwise { kernel: Add }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 256, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 256, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 256, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 256, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 256, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 256, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0, 1]) }, deferred_output: None }, format_policy: RowMajorGrid }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: LayerNorm, dispatch: Pointwise { kernel: LayerNorm }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 256, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 256, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 256, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 256, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 256, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 256, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 256, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 256, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: Fresh }, deferred_output: None }, format_policy: RowMajorRows }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: LayerNorm, dispatch: Pointwise { kernel: LayerNorm }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 256, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 256, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 256, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 256, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 256, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 256, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 256, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 256, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: Fresh }, deferred_output: None }, format_policy: Concrete }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Gelu, dispatch: Pointwise { kernel: Gelu }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: Amp(Left), tiling: TensorTiling { tile_count: 256, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 256, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }, AxisTiling { axis: FromEnd(1), partitions: 1, padding_groups: 1, block_size: 64, padding_multiple: 64, shard_padding_multiple: 1, padding: Zero, tile_stride: None }] }, memory_class: Ipu21Interleaved } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: Amp(Left), tiling: TensorTiling { tile_count: 256, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 256, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }, AxisTiling { axis: FromEnd(1), partitions: 1, padding_groups: 1, block_size: 64, padding_multiple: 64, shard_padding_multiple: 1, padding: Zero, tile_stride: None }] }, memory_class: Ipu21Interleaved } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0]) }, deferred_output: None }, format_policy: PreserveInputLayout(0) }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Gelu, dispatch: Pointwise { kernel: Gelu }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 256, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 256, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 256, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 256, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0]) }, deferred_output: None }, format_policy: PreserveInputLayout(0) }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Add, dispatch: Pointwise { kernel: Add }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 256, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 256, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 256, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 256, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 256, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 256, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0, 1]) }, deferred_output: None }, format_policy: Concrete }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: FlashAttention { options: AttentionOptions { causal: false, scale: InverseSqrtQueryWidth }, accumulate: F32 }, dispatch: Pointwise { kernel: FlashAttention { options: AttentionOptions { causal: false, scale: InverseSqrtQueryWidth }, accumulate: F32 } }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 256, replicas: 1, axes: [AxisTiling { axis: FromEnd(3), partitions: 256, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 256, replicas: 1, axes: [AxisTiling { axis: FromEnd(3), partitions: 256, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 256, replicas: 1, axes: [AxisTiling { axis: FromEnd(3), partitions: 256, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 256, replicas: 1, axes: [AxisTiling { axis: FromEnd(3), partitions: 256, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: Fresh }, deferred_output: None }, format_policy: Concrete }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Add, dispatch: Pointwise { kernel: Add }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 128, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 128, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 128, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 128, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 128, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 128, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0, 1]) }, deferred_output: None }, format_policy: PreserveInputLayout(0) }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Add, dispatch: Pointwise { kernel: Add }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 128, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 128, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 128, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 128, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 128, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 128, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0, 1]) }, deferred_output: None }, format_policy: PreserveInputLayout(1) }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Add, dispatch: Pointwise { kernel: Add }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 128, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 128, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 128, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 128, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 128, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 128, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0, 1]) }, deferred_output: None }, format_policy: RowMajorGrid }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: LayerNorm, dispatch: Pointwise { kernel: LayerNorm }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 128, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 128, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 128, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 128, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 128, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 128, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 128, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 128, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: Fresh }, deferred_output: None }, format_policy: RowMajorRows }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: LayerNorm, dispatch: Pointwise { kernel: LayerNorm }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 128, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 128, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 128, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 128, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 128, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 128, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 128, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 128, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: Fresh }, deferred_output: None }, format_policy: Concrete }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Gelu, dispatch: Pointwise { kernel: Gelu }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: Amp(Left), tiling: TensorTiling { tile_count: 128, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 128, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }, AxisTiling { axis: FromEnd(1), partitions: 1, padding_groups: 1, block_size: 64, padding_multiple: 64, shard_padding_multiple: 1, padding: Zero, tile_stride: None }] }, memory_class: Ipu21Interleaved } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: Amp(Left), tiling: TensorTiling { tile_count: 128, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 128, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }, AxisTiling { axis: FromEnd(1), partitions: 1, padding_groups: 1, block_size: 64, padding_multiple: 64, shard_padding_multiple: 1, padding: Zero, tile_stride: None }] }, memory_class: Ipu21Interleaved } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0]) }, deferred_output: None }, format_policy: PreserveInputLayout(0) }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Gelu, dispatch: Pointwise { kernel: Gelu }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 128, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 128, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 128, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 128, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0]) }, deferred_output: None }, format_policy: PreserveInputLayout(0) }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Add, dispatch: Pointwise { kernel: Add }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 128, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 128, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 128, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 128, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 128, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 128, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0, 1]) }, deferred_output: None }, format_policy: Concrete }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: FlashAttention { options: AttentionOptions { causal: false, scale: InverseSqrtQueryWidth }, accumulate: F32 }, dispatch: Pointwise { kernel: FlashAttention { options: AttentionOptions { causal: false, scale: InverseSqrtQueryWidth }, accumulate: F32 } }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 128, replicas: 1, axes: [AxisTiling { axis: FromEnd(3), partitions: 128, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 128, replicas: 1, axes: [AxisTiling { axis: FromEnd(3), partitions: 128, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 128, replicas: 1, axes: [AxisTiling { axis: FromEnd(3), partitions: 128, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 128, replicas: 1, axes: [AxisTiling { axis: FromEnd(3), partitions: 128, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: Fresh }, deferred_output: None }, format_policy: Concrete }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Add, dispatch: Pointwise { kernel: Add }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 64, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 64, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 64, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 64, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 64, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 64, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0, 1]) }, deferred_output: None }, format_policy: PreserveInputLayout(0) }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Add, dispatch: Pointwise { kernel: Add }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 64, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 64, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 64, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 64, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 64, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 64, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0, 1]) }, deferred_output: None }, format_policy: PreserveInputLayout(1) }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Add, dispatch: Pointwise { kernel: Add }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 64, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 64, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 64, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 64, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 64, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 64, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0, 1]) }, deferred_output: None }, format_policy: RowMajorGrid }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: LayerNorm, dispatch: Pointwise { kernel: LayerNorm }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 64, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 64, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 64, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 64, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 64, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 64, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 64, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 64, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: Fresh }, deferred_output: None }, format_policy: RowMajorRows }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: LayerNorm, dispatch: Pointwise { kernel: LayerNorm }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 64, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 64, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 64, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 64, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 64, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 64, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 64, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 64, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: Fresh }, deferred_output: None }, format_policy: Concrete }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Gelu, dispatch: Pointwise { kernel: Gelu }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: Amp(Left), tiling: TensorTiling { tile_count: 64, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 64, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }, AxisTiling { axis: FromEnd(1), partitions: 1, padding_groups: 1, block_size: 64, padding_multiple: 64, shard_padding_multiple: 1, padding: Zero, tile_stride: None }] }, memory_class: Ipu21Interleaved } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: Amp(Left), tiling: TensorTiling { tile_count: 64, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 64, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }, AxisTiling { axis: FromEnd(1), partitions: 1, padding_groups: 1, block_size: 64, padding_multiple: 64, shard_padding_multiple: 1, padding: Zero, tile_stride: None }] }, memory_class: Ipu21Interleaved } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0]) }, deferred_output: None }, format_policy: PreserveInputLayout(0) }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Gelu, dispatch: Pointwise { kernel: Gelu }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 64, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 64, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 64, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 64, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0]) }, deferred_output: None }, format_policy: PreserveInputLayout(0) }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Add, dispatch: Pointwise { kernel: Add }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 64, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 64, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 64, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 64, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 64, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 64, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0, 1]) }, deferred_output: None }, format_policy: Concrete }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: FlashAttention { options: AttentionOptions { causal: false, scale: InverseSqrtQueryWidth }, accumulate: F32 }, dispatch: Pointwise { kernel: FlashAttention { options: AttentionOptions { causal: false, scale: InverseSqrtQueryWidth }, accumulate: F32 } }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 64, replicas: 1, axes: [AxisTiling { axis: FromEnd(3), partitions: 64, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 64, replicas: 1, axes: [AxisTiling { axis: FromEnd(3), partitions: 64, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 64, replicas: 1, axes: [AxisTiling { axis: FromEnd(3), partitions: 64, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 64, replicas: 1, axes: [AxisTiling { axis: FromEnd(3), partitions: 64, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: Fresh }, deferred_output: None }, format_policy: Concrete }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Add, dispatch: Pointwise { kernel: Add }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 32, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 32, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 32, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 32, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 32, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 32, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0, 1]) }, deferred_output: None }, format_policy: PreserveInputLayout(0) }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Add, dispatch: Pointwise { kernel: Add }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 32, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 32, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 32, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 32, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 32, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 32, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0, 1]) }, deferred_output: None }, format_policy: PreserveInputLayout(1) }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Add, dispatch: Pointwise { kernel: Add }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 32, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 32, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 32, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 32, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 32, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 32, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0, 1]) }, deferred_output: None }, format_policy: RowMajorGrid }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: LayerNorm, dispatch: Pointwise { kernel: LayerNorm }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 32, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 32, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 32, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 32, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 32, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 32, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 32, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 32, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: Fresh }, deferred_output: None }, format_policy: RowMajorRows }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: LayerNorm, dispatch: Pointwise { kernel: LayerNorm }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 32, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 32, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 32, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 32, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 32, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 32, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 32, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 32, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: Fresh }, deferred_output: None }, format_policy: Concrete }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Gelu, dispatch: Pointwise { kernel: Gelu }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: Amp(Left), tiling: TensorTiling { tile_count: 32, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 32, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }, AxisTiling { axis: FromEnd(1), partitions: 1, padding_groups: 1, block_size: 64, padding_multiple: 64, shard_padding_multiple: 1, padding: Zero, tile_stride: None }] }, memory_class: Ipu21Interleaved } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: Amp(Left), tiling: TensorTiling { tile_count: 32, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 32, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }, AxisTiling { axis: FromEnd(1), partitions: 1, padding_groups: 1, block_size: 64, padding_multiple: 64, shard_padding_multiple: 1, padding: Zero, tile_stride: None }] }, memory_class: Ipu21Interleaved } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0]) }, deferred_output: None }, format_policy: PreserveInputLayout(0) }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Gelu, dispatch: Pointwise { kernel: Gelu }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 32, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 32, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 32, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 32, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0]) }, deferred_output: None }, format_policy: PreserveInputLayout(0) }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Add, dispatch: Pointwise { kernel: Add }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 32, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 32, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 32, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 32, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 32, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 32, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0, 1]) }, deferred_output: None }, format_policy: Concrete }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: FlashAttention { options: AttentionOptions { causal: false, scale: InverseSqrtQueryWidth }, accumulate: F32 }, dispatch: Pointwise { kernel: FlashAttention { options: AttentionOptions { causal: false, scale: InverseSqrtQueryWidth }, accumulate: F32 } }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 32, replicas: 1, axes: [AxisTiling { axis: FromEnd(3), partitions: 32, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 32, replicas: 1, axes: [AxisTiling { axis: FromEnd(3), partitions: 32, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 32, replicas: 1, axes: [AxisTiling { axis: FromEnd(3), partitions: 32, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 32, replicas: 1, axes: [AxisTiling { axis: FromEnd(3), partitions: 32, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: Fresh }, deferred_output: None }, format_policy: Concrete }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Add, dispatch: Pointwise { kernel: Add }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 16, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 16, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 16, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 16, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 16, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 16, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0, 1]) }, deferred_output: None }, format_policy: PreserveInputLayout(0) }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Add, dispatch: Pointwise { kernel: Add }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 16, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 16, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 16, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 16, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 16, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 16, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0, 1]) }, deferred_output: None }, format_policy: PreserveInputLayout(1) }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Add, dispatch: Pointwise { kernel: Add }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 16, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 16, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 16, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 16, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 16, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 16, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0, 1]) }, deferred_output: None }, format_policy: RowMajorGrid }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: LayerNorm, dispatch: Pointwise { kernel: LayerNorm }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 16, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 16, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 16, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 16, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 16, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 16, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 16, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 16, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: Fresh }, deferred_output: None }, format_policy: RowMajorRows }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: LayerNorm, dispatch: Pointwise { kernel: LayerNorm }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 16, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 16, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 16, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 16, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 16, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 16, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 16, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 16, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: Fresh }, deferred_output: None }, format_policy: Concrete }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Gelu, dispatch: Pointwise { kernel: Gelu }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: Amp(Left), tiling: TensorTiling { tile_count: 16, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 16, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }, AxisTiling { axis: FromEnd(1), partitions: 1, padding_groups: 1, block_size: 64, padding_multiple: 64, shard_padding_multiple: 1, padding: Zero, tile_stride: None }] }, memory_class: Ipu21Interleaved } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: Amp(Left), tiling: TensorTiling { tile_count: 16, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 16, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }, AxisTiling { axis: FromEnd(1), partitions: 1, padding_groups: 1, block_size: 64, padding_multiple: 64, shard_padding_multiple: 1, padding: Zero, tile_stride: None }] }, memory_class: Ipu21Interleaved } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0]) }, deferred_output: None }, format_policy: PreserveInputLayout(0) }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Gelu, dispatch: Pointwise { kernel: Gelu }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 16, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 16, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 16, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 16, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0]) }, deferred_output: None }, format_policy: PreserveInputLayout(0) }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Add, dispatch: Pointwise { kernel: Add }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 16, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 16, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 16, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 16, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 16, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 16, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0, 1]) }, deferred_output: None }, format_policy: Concrete }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: FlashAttention { options: AttentionOptions { causal: false, scale: InverseSqrtQueryWidth }, accumulate: F32 }, dispatch: Pointwise { kernel: FlashAttention { options: AttentionOptions { causal: false, scale: InverseSqrtQueryWidth }, accumulate: F32 } }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 16, replicas: 1, axes: [AxisTiling { axis: FromEnd(3), partitions: 16, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 16, replicas: 1, axes: [AxisTiling { axis: FromEnd(3), partitions: 16, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 16, replicas: 1, axes: [AxisTiling { axis: FromEnd(3), partitions: 16, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 16, replicas: 1, axes: [AxisTiling { axis: FromEnd(3), partitions: 16, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: Fresh }, deferred_output: None }, format_policy: Concrete }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Add, dispatch: Pointwise { kernel: Add }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 8, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 8, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 8, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 8, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 8, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 8, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0, 1]) }, deferred_output: None }, format_policy: PreserveInputLayout(0) }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Add, dispatch: Pointwise { kernel: Add }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 8, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 8, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 8, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 8, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 8, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 8, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0, 1]) }, deferred_output: None }, format_policy: PreserveInputLayout(1) }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Add, dispatch: Pointwise { kernel: Add }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 8, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 8, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 8, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 8, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 8, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 8, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0, 1]) }, deferred_output: None }, format_policy: RowMajorGrid }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: LayerNorm, dispatch: Pointwise { kernel: LayerNorm }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 8, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 8, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 8, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 8, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 8, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 8, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 8, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 8, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: Fresh }, deferred_output: None }, format_policy: RowMajorRows }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: LayerNorm, dispatch: Pointwise { kernel: LayerNorm }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 8, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 8, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 8, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 8, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 8, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 8, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 8, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 8, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: Fresh }, deferred_output: None }, format_policy: Concrete }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Gelu, dispatch: Pointwise { kernel: Gelu }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: Amp(Left), tiling: TensorTiling { tile_count: 8, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 8, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }, AxisTiling { axis: FromEnd(1), partitions: 1, padding_groups: 1, block_size: 64, padding_multiple: 64, shard_padding_multiple: 1, padding: Zero, tile_stride: None }] }, memory_class: Ipu21Interleaved } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: Amp(Left), tiling: TensorTiling { tile_count: 8, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 8, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }, AxisTiling { axis: FromEnd(1), partitions: 1, padding_groups: 1, block_size: 64, padding_multiple: 64, shard_padding_multiple: 1, padding: Zero, tile_stride: None }] }, memory_class: Ipu21Interleaved } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0]) }, deferred_output: None }, format_policy: PreserveInputLayout(0) }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Gelu, dispatch: Pointwise { kernel: Gelu }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 8, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 8, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 8, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 8, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0]) }, deferred_output: None }, format_policy: PreserveInputLayout(0) }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Add, dispatch: Pointwise { kernel: Add }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 8, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 8, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 8, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 8, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 8, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 8, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0, 1]) }, deferred_output: None }, format_policy: Concrete }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: FlashAttention { options: AttentionOptions { causal: false, scale: InverseSqrtQueryWidth }, accumulate: F32 }, dispatch: Pointwise { kernel: FlashAttention { options: AttentionOptions { causal: false, scale: InverseSqrtQueryWidth }, accumulate: F32 } }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 8, replicas: 1, axes: [AxisTiling { axis: FromEnd(3), partitions: 8, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 8, replicas: 1, axes: [AxisTiling { axis: FromEnd(3), partitions: 8, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 8, replicas: 1, axes: [AxisTiling { axis: FromEnd(3), partitions: 8, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 8, replicas: 1, axes: [AxisTiling { axis: FromEnd(3), partitions: 8, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: Fresh }, deferred_output: None }, format_policy: Concrete }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Add, dispatch: Pointwise { kernel: Add }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 4, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 4, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 4, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 4, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 4, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 4, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0, 1]) }, deferred_output: None }, format_policy: PreserveInputLayout(0) }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Add, dispatch: Pointwise { kernel: Add }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 4, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 4, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 4, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 4, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 4, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 4, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0, 1]) }, deferred_output: None }, format_policy: PreserveInputLayout(1) }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Add, dispatch: Pointwise { kernel: Add }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 4, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 4, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 4, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 4, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 4, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 4, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0, 1]) }, deferred_output: None }, format_policy: RowMajorGrid }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: LayerNorm, dispatch: Pointwise { kernel: LayerNorm }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 4, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 4, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 4, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 4, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 4, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 4, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 4, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 4, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: Fresh }, deferred_output: None }, format_policy: RowMajorRows }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: LayerNorm, dispatch: Pointwise { kernel: LayerNorm }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 4, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 4, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 4, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 4, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 4, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 4, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 4, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 4, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: Fresh }, deferred_output: None }, format_policy: Concrete }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Gelu, dispatch: Pointwise { kernel: Gelu }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: Amp(Left), tiling: TensorTiling { tile_count: 4, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 4, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }, AxisTiling { axis: FromEnd(1), partitions: 1, padding_groups: 1, block_size: 64, padding_multiple: 64, shard_padding_multiple: 1, padding: Zero, tile_stride: None }] }, memory_class: Ipu21Interleaved } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: Amp(Left), tiling: TensorTiling { tile_count: 4, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 4, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }, AxisTiling { axis: FromEnd(1), partitions: 1, padding_groups: 1, block_size: 64, padding_multiple: 64, shard_padding_multiple: 1, padding: Zero, tile_stride: None }] }, memory_class: Ipu21Interleaved } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0]) }, deferred_output: None }, format_policy: PreserveInputLayout(0) }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Gelu, dispatch: Pointwise { kernel: Gelu }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 4, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 4, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 4, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 4, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0]) }, deferred_output: None }, format_policy: PreserveInputLayout(0) }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Add, dispatch: Pointwise { kernel: Add }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 4, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 4, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 4, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 4, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 4, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 4, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0, 1]) }, deferred_output: None }, format_policy: Concrete }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: FlashAttention { options: AttentionOptions { causal: false, scale: InverseSqrtQueryWidth }, accumulate: F32 }, dispatch: Pointwise { kernel: FlashAttention { options: AttentionOptions { causal: false, scale: InverseSqrtQueryWidth }, accumulate: F32 } }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 4, replicas: 1, axes: [AxisTiling { axis: FromEnd(3), partitions: 4, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 4, replicas: 1, axes: [AxisTiling { axis: FromEnd(3), partitions: 4, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 4, replicas: 1, axes: [AxisTiling { axis: FromEnd(3), partitions: 4, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 4, replicas: 1, axes: [AxisTiling { axis: FromEnd(3), partitions: 4, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: Fresh }, deferred_output: None }, format_policy: Concrete }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Add, dispatch: Pointwise { kernel: Add }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 2, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 2, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 2, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 2, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 2, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 2, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0, 1]) }, deferred_output: None }, format_policy: PreserveInputLayout(0) }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Add, dispatch: Pointwise { kernel: Add }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 2, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 2, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 2, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 2, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 2, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 2, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0, 1]) }, deferred_output: None }, format_policy: PreserveInputLayout(1) }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Add, dispatch: Pointwise { kernel: Add }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 2, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 2, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 2, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 2, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 2, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 2, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0, 1]) }, deferred_output: None }, format_policy: RowMajorGrid }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: LayerNorm, dispatch: Pointwise { kernel: LayerNorm }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 2, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 2, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 2, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 2, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 2, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 2, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 2, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 2, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: Fresh }, deferred_output: None }, format_policy: RowMajorRows }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: LayerNorm, dispatch: Pointwise { kernel: LayerNorm }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 2, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 2, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 2, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 2, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 2, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 2, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 2, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 2, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: Fresh }, deferred_output: None }, format_policy: Concrete }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Gelu, dispatch: Pointwise { kernel: Gelu }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: Amp(Left), tiling: TensorTiling { tile_count: 2, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 2, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }, AxisTiling { axis: FromEnd(1), partitions: 1, padding_groups: 1, block_size: 64, padding_multiple: 64, shard_padding_multiple: 1, padding: Zero, tile_stride: None }] }, memory_class: Ipu21Interleaved } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: Amp(Left), tiling: TensorTiling { tile_count: 2, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 2, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }, AxisTiling { axis: FromEnd(1), partitions: 1, padding_groups: 1, block_size: 64, padding_multiple: 64, shard_padding_multiple: 1, padding: Zero, tile_stride: None }] }, memory_class: Ipu21Interleaved } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0]) }, deferred_output: None }, format_policy: PreserveInputLayout(0) }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Gelu, dispatch: Pointwise { kernel: Gelu }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 2, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 2, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 2, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 2, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0]) }, deferred_output: None }, format_policy: PreserveInputLayout(0) }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Add, dispatch: Pointwise { kernel: Add }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 2, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 2, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 2, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 2, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 2, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 2, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0, 1]) }, deferred_output: None }, format_policy: Concrete }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: FlashAttention { options: AttentionOptions { causal: false, scale: InverseSqrtQueryWidth }, accumulate: F32 }, dispatch: Pointwise { kernel: FlashAttention { options: AttentionOptions { causal: false, scale: InverseSqrtQueryWidth }, accumulate: F32 } }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 2, replicas: 1, axes: [AxisTiling { axis: FromEnd(3), partitions: 2, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 2, replicas: 1, axes: [AxisTiling { axis: FromEnd(3), partitions: 2, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 2, replicas: 1, axes: [AxisTiling { axis: FromEnd(3), partitions: 2, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 2, replicas: 1, axes: [AxisTiling { axis: FromEnd(3), partitions: 2, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: Fresh }, deferred_output: None }, format_policy: Concrete }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Add, dispatch: Pointwise { kernel: Add }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0, 1]) }, deferred_output: None }, format_policy: PreserveInputLayout(0) }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Add, dispatch: Pointwise { kernel: Add }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0, 1]) }, deferred_output: None }, format_policy: PreserveInputLayout(1) }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Add, dispatch: Pointwise { kernel: Add }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0, 1]) }, deferred_output: None }, format_policy: RowMajorGrid }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: LayerNorm, dispatch: Pointwise { kernel: LayerNorm }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: Fresh }, deferred_output: None }, format_policy: RowMajorRows }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: LayerNorm, dispatch: Pointwise { kernel: LayerNorm }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: Fresh }, deferred_output: None }, format_policy: Concrete }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Gelu, dispatch: Pointwise { kernel: Gelu }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: Amp(Left), tiling: TensorTiling { tile_count: 1, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }, AxisTiling { axis: FromEnd(1), partitions: 1, padding_groups: 1, block_size: 64, padding_multiple: 64, shard_padding_multiple: 1, padding: Zero, tile_stride: None }] }, memory_class: Ipu21Interleaved } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: Amp(Left), tiling: TensorTiling { tile_count: 1, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }, AxisTiling { axis: FromEnd(1), partitions: 1, padding_groups: 1, block_size: 64, padding_multiple: 64, shard_padding_multiple: 1, padding: Zero, tile_stride: None }] }, memory_class: Ipu21Interleaved } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0]) }, deferred_output: None }, format_policy: PreserveInputLayout(0) }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Gelu, dispatch: Pointwise { kernel: Gelu }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0]) }, deferred_output: None }, format_policy: PreserveInputLayout(0) }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: Add, dispatch: Pointwise { kernel: Add }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1, replicas: 1, axes: [AxisTiling { axis: FromEnd(2), partitions: 1, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: MayAliasInputs([0, 1]) }, deferred_output: None }, format_policy: Concrete }), Concrete(ConcreteOperatorCandidate { plan: OperatorPlan { operator: FlashAttention { options: AttentionOptions { causal: false, scale: InverseSqrtQueryWidth }, accumulate: F32 }, dispatch: Pointwise { kernel: FlashAttention { options: AttentionOptions { causal: false, scale: InverseSqrtQueryWidth }, accumulate: F32 } }, requirements: StorageRequirements { inputs: [OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1, replicas: 1, axes: [AxisTiling { axis: FromEnd(3), partitions: 1, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1, replicas: 1, axes: [AxisTiling { axis: FromEnd(3), partitions: 1, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1, replicas: 1, axes: [AxisTiling { axis: FromEnd(3), partitions: 1, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }], output: OperandRequirement { format: TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 1, replicas: 1, axes: [AxisTiling { axis: FromEnd(3), partitions: 1, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: None }] }, memory_class: Ipu21Standard } }, local_staging: Direct, materialization: Complete }, output_aliasing: Fresh }, deferred_output: None }, format_policy: Concrete }), ParallelGemm { tile_count: 1472, multiply: F8F143 { scale_exponent: -4 }, options: GemmOptions { transpose_left: false, transpose_right: false }, accumulate: F16 }], gemm_precisions: {}, shape_aware_active_tile_counts: true, operator_candidate_limit: 64, exchange_table_budget_bytes: 98304, exchange_transfer_limit_per_tile: 16384, exchange_table_cost_per_byte: 0, gemm_plan_constraints: [], gemm_output_packing: Automatic, max_parallel_reductions: 3, standard_memory_reservation_bytes: 49152, memory_profile_directory: None, tile_memory_budget_bytes: 586960, profiling: true, diagnostic_checkpoints: false, exchange_diagnostics: false, conversion_streaming: WhenRequired, attention_strategy: Automatic, attention_products: Automatic, attention_fp8_scales: [None, None] }\nNone", "recipe": {"plans": {"0": {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 6, "column_partitions": 24, "inner_partitions": 10, "result_row_partitions": 10, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 24, "axes": [{"axis": {"FromEnd": 1}, "partitions": 10, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 240, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 60, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 24}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, "1": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, "2": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, "3": {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1458, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1458, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, "4": {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 96, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 36, "inner_partitions": 4, "result_row_partitions": 1, "result_column_partitions": 4, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 144}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, "5": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1458, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, "6": {"operator": {"Slice": {"axis": 2, "start": 0, "length": 1152}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, "7": {"operator": {"Slice": {"axis": 2, "start": 1152, "length": 1152}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, "8": {"operator": {"Slice": {"axis": 2, "start": 2304, "length": 1152}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, "9": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 16, "reversed": false}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": {"source_input": 0, "transform": {"split_axis": 2, "merge_axis": 0, "factor": 16, "reversed": false}, "unfused_cycles": 0, "unfused_exchange_cycles": 0}}, "10": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 16, "reversed": false}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": {"source_input": 0, "transform": {"split_axis": 2, "merge_axis": 0, "factor": 16, "reversed": false}, "unfused_cycles": 0, "unfused_exchange_cycles": 0}}, "11": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 16, "reversed": false}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": {"source_input": 0, "transform": {"split_axis": 2, "merge_axis": 0, "factor": 16, "reversed": false}, "unfused_cycles": 0, "unfused_exchange_cycles": 0}}, "12": {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, "13": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 16, "reversed": true}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, "14": {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 576, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 61, "column_partitions": 2, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 12, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1464, "replicas": 2, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 61, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 24, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1464, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 61, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 24}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, "15": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1458, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, "16": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}, "17": {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1458, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1458, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, "18": {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 80, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 9, "column_partitions": 54, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 54, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 162, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 27, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 54}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, "19": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1458, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, "20": {"operator": "Gelu", "dispatch": {"Pointwise": {"kernel": "Gelu"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, "21": {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 192, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 6, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 12, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, "22": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1458, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, "23": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}, "25": {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1458, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1458, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, "26": {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 96, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 24, "inner_partitions": 6, "result_row_partitions": 1, "result_column_partitions": 6, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 24, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 144}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, "27": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, "28": {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 18, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, "29": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 576, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 288}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 288}, {"axis": {"FromEnd": 1}, "partitions": 288, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 576, "replicas": 2, "axes": [{"axis": {"FromStart": 2}, "partitions": 288, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 576, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 288}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 288}, {"axis": {"FromEnd": 1}, "partitions": 288, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, "30": {"operator": {"Slice": {"axis": 2, "start": 0, "length": 1152}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, "31": {"operator": {"Slice": {"axis": 2, "start": 1152, "length": 1152}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, "32": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 16, "reversed": false}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": {"source_input": 0, "transform": {"split_axis": 2, "merge_axis": 0, "factor": 16, "reversed": false}, "unfused_cycles": 0, "unfused_exchange_cycles": 0}}, "33": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 16, "reversed": false}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": {"source_input": 0, "transform": {"split_axis": 2, "merge_axis": 0, "factor": 16, "reversed": false}, "unfused_cycles": 0, "unfused_exchange_cycles": 0}}, "34": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 16, "reversed": false}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": {"source_input": 0, "transform": {"split_axis": 2, "merge_axis": 0, "factor": 16, "reversed": false}, "unfused_cycles": 0, "unfused_exchange_cycles": 0}}, "35": {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, "36": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 16, "reversed": true}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, "37": {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 18, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, "38": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 576, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 288}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 288}, {"axis": {"FromEnd": 1}, "partitions": 288, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 576, "replicas": 2, "axes": [{"axis": {"FromStart": 2}, "partitions": 288, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 576, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 288}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 288}, {"axis": {"FromEnd": 1}, "partitions": 288, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, "39": {"operator": "LayerNorm", "dispatch": {"LayerNorm": {"parts": 4}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 4}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 4}, {"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 2, "axes": [{"axis": {"FromStart": 2}, "partitions": 4, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 2, "axes": [{"axis": {"FromStart": 2}, "partitions": 4, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 4}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 4}, {"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, "40": {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 184, "inner_partitions": 8, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 184, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 184, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 184}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, "41": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 736}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 736}, {"axis": {"FromEnd": 1}, "partitions": 736, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 2, "axes": [{"axis": {"FromStart": 2}, "partitions": 736, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 736}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 736}, {"axis": {"FromEnd": 1}, "partitions": 736, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, "42": {"operator": "Gelu", "dispatch": {"Pointwise": {"kernel": "Gelu"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 16, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 16, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, "43": {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 64, "inner_partitions": 23, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 64, "axes": [{"axis": {"FromEnd": 1}, "partitions": 23, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 64, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 23, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 64, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 64}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, "44": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 576, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 288}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 288}, {"axis": {"FromEnd": 1}, "partitions": 288, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 576, "replicas": 2, "axes": [{"axis": {"FromStart": 2}, "partitions": 288, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 576, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 288}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 288}, {"axis": {"FromEnd": 1}, "partitions": 288, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, "45": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}}, "open_boundaries": [346, 350, 353, 354, 355, 360, 362, 391, 396], "early_casts": [0, 4, 14, 18, 21, 26, 40], "packing_rows": 128, "parallel_reductions": 0, "disjoint_copy_sources": false, "in_place_casts": null, "cast_before_copies": [[12, 1]]}, "alternatives": {"0": [{"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 72, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 72, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 72, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 72, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 72, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 72, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 72, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 72, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 72, "inner_partitions": 4, "result_row_partitions": 4, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 72, "inner_partitions": 4, "result_row_partitions": 4, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 224, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 6, "column_partitions": 72, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 224, "column_block": 16}}}, "tiling": {"tile_count": 216, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 224, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 6, "column_partitions": 72, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 224, "column_block": 16}}}, "tiling": {"tile_count": 216, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 72, "inner_partitions": 4, "result_row_partitions": 4, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 72, "inner_partitions": 4, "result_row_partitions": 4, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 224, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 6, "column_partitions": 72, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 224, "column_block": 16}}}, "tiling": {"tile_count": 216, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 224, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 6, "column_partitions": 72, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 224, "column_block": 16}}}, "tiling": {"tile_count": 216, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 4, "column_partitions": 72, "inner_partitions": 5, "result_row_partitions": 5, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 360, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 4, "column_partitions": 72, "inner_partitions": 5, "result_row_partitions": 5, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 360, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 72, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 72, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 72, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 72, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 72, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 10}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 72, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 10}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 72, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 10}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 72, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 10}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 36, "inner_partitions": 4, "result_row_partitions": 2, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 36, "inner_partitions": 4, "result_row_partitions": 2, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 36, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 36, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 36, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 2, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 36, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 2, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 4, "column_partitions": 72, "inner_partitions": 5, "result_row_partitions": 5, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 360, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 4, "column_partitions": 72, "inner_partitions": 5, "result_row_partitions": 5, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 360, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 36, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 40, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 36}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 36, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 40, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 36}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 36, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 40, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 36}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 36, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 40, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 36}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 36, "inner_partitions": 4, "result_row_partitions": 4, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 40, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 36}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 36, "inner_partitions": 4, "result_row_partitions": 4, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 40, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 36}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 224, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 13, "column_partitions": 37, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1443, "replicas": 37, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 13, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 224, "column_block": 16}}}, "tiling": {"tile_count": 111, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 37, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1443, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 37, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 39, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 37}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 224, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 13, "column_partitions": 37, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1443, "replicas": 37, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 13, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 224, "column_block": 16}}}, "tiling": {"tile_count": 111, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 37, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1443, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 37, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 39, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 37}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 72, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 72, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 72, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 72, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 72, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 10}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 72, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 10}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 72, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 10}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 72, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 10}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 36, "inner_partitions": 4, "result_row_partitions": 2, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 36, "inner_partitions": 4, "result_row_partitions": 2, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 224, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 6, "column_partitions": 72, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 224, "column_block": 16}}}, "tiling": {"tile_count": 216, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 224, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 6, "column_partitions": 72, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 224, "column_block": 16}}}, "tiling": {"tile_count": 216, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 36, "inner_partitions": 4, "result_row_partitions": 4, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 40, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 36}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 36, "inner_partitions": 4, "result_row_partitions": 4, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 40, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 36}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 8, "column_partitions": 36, "inner_partitions": 5, "result_row_partitions": 5, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 180, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 40, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 36}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 8, "column_partitions": 36, "inner_partitions": 5, "result_row_partitions": 5, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 180, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 40, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 36}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 224, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 13, "column_partitions": 37, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1443, "replicas": 37, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 13, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 224, "column_block": 16}}}, "tiling": {"tile_count": 111, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 37, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1443, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 37, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 39, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 37}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 224, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 13, "column_partitions": 37, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1443, "replicas": 37, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 13, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 224, "column_block": 16}}}, "tiling": {"tile_count": 111, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 37, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1443, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 37, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 39, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 37}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 224, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 6, "column_partitions": 72, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 224, "column_block": 16}}}, "tiling": {"tile_count": 216, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 224, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 6, "column_partitions": 72, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 224, "column_block": 16}}}, "tiling": {"tile_count": 216, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 36, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 36, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 36, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 2, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 36, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 2, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 36, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 40, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 36}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 36, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 40, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 36}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 36, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 40, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 36}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 36, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 40, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 36}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 8, "column_partitions": 36, "inner_partitions": 5, "result_row_partitions": 5, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 180, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 40, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 36}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 8, "column_partitions": 36, "inner_partitions": 5, "result_row_partitions": 5, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 180, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 40, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 36}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 224, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 24, "inner_partitions": 3, "result_row_partitions": 1, "result_column_partitions": 3, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 24, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 224, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 224, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 24, "inner_partitions": 3, "result_row_partitions": 1, "result_column_partitions": 3, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 24, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 224, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 15, "column_partitions": 24, "inner_partitions": 4, "result_row_partitions": 4, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 24, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 15, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 96, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 60, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 24}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 15, "column_partitions": 24, "inner_partitions": 4, "result_row_partitions": 4, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 24, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 15, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 96, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 60, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 24}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 224, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 13, "column_partitions": 37, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1443, "replicas": 37, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 13, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 224, "column_block": 16}}}, "tiling": {"tile_count": 111, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 37, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1443, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 37, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 39, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 37}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 224, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 13, "column_partitions": 37, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1443, "replicas": 37, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 13, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 224, "column_block": 16}}}, "tiling": {"tile_count": 111, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 37, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1443, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 37, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 39, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 37}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 12, "column_partitions": 24, "inner_partitions": 5, "result_row_partitions": 5, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 24, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 120, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 60, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 24}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 12, "column_partitions": 24, "inner_partitions": 5, "result_row_partitions": 5, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 24, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 120, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 60, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 24}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 72, "inner_partitions": 4, "result_row_partitions": 4, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 72, "inner_partitions": 4, "result_row_partitions": 4, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 36, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 36}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 36, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 36}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 36, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 36}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 36, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 36}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 36, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 20}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 36, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 20}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 36, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 20}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 36, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 20}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 18, "inner_partitions": 4, "result_row_partitions": 1, "result_column_partitions": 4, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 18, "inner_partitions": 4, "result_row_partitions": 1, "result_column_partitions": 4, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 18, "inner_partitions": 4, "result_row_partitions": 1, "result_column_partitions": 4, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 18, "inner_partitions": 4, "result_row_partitions": 1, "result_column_partitions": 4, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 72, "inner_partitions": 4, "result_row_partitions": 4, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 72, "inner_partitions": 4, "result_row_partitions": 4, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 18, "inner_partitions": 4, "result_row_partitions": 2, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 40, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 36}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 18, "inner_partitions": 4, "result_row_partitions": 2, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 40, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 36}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 8, "column_partitions": 26, "inner_partitions": 7, "result_row_partitions": 7, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1456, "replicas": 26, "axes": [{"axis": {"FromEnd": 1}, "partitions": 7, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 182, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 26, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 7, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1456, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 26, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 56, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 26}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 8, "column_partitions": 26, "inner_partitions": 7, "result_row_partitions": 7, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1456, "replicas": 26, "axes": [{"axis": {"FromEnd": 1}, "partitions": 7, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 182, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 26, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 7, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1456, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 26, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 56, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 26}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 72, "inner_partitions": 10, "result_row_partitions": 10, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 10, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 72, "inner_partitions": 10, "result_row_partitions": 10, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 10, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 4, "column_partitions": 36, "inner_partitions": 10, "result_row_partitions": 5, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 10, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 360, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 4, "column_partitions": 36, "inner_partitions": 10, "result_row_partitions": 5, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 10, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 360, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 224, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 6, "column_partitions": 72, "inner_partitions": 3, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 224, "column_block": 16}}}, "tiling": {"tile_count": 216, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 432, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 224, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 6, "column_partitions": 72, "inner_partitions": 3, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 224, "column_block": 16}}}, "tiling": {"tile_count": 216, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 432, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 128, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 32, "column_partitions": 9, "inner_partitions": 5, "result_row_partitions": 1, "result_column_partitions": 5, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 45, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 45, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 45}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 128, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 32, "column_partitions": 9, "inner_partitions": 5, "result_row_partitions": 1, "result_column_partitions": 5, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 45, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 45, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 45}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 192, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 35, "column_partitions": 6, "inner_partitions": 7, "result_row_partitions": 1, "result_column_partitions": 7, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 7, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 35, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 42, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 7, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 42, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 35, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 42}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 192, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 35, "column_partitions": 6, "inner_partitions": 7, "result_row_partitions": 1, "result_column_partitions": 7, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 7, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 35, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 42, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 7, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 42, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 35, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 42}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 72, "inner_partitions": 4, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 360, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 72, "inner_partitions": 4, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 360, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 19, "result_row_partitions": 19, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1368, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 19, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 1368, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 19, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1368, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 19, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 19, "result_row_partitions": 19, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1368, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 19, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 1368, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 19, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1368, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 19, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 384, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 49, "column_partitions": 3, "inner_partitions": 10, "result_row_partitions": 1, "result_column_partitions": 10, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 3, "axes": [{"axis": {"FromEnd": 1}, "partitions": 10, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 49, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 30, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 30, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 49, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 30}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 384, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 49, "column_partitions": 3, "inner_partitions": 10, "result_row_partitions": 1, "result_column_partitions": 10, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 3, "axes": [{"axis": {"FromEnd": 1}, "partitions": 10, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 49, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 30, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 30, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 49, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 30}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 6, "column_partitions": 24, "inner_partitions": 10, "result_row_partitions": 10, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 24, "axes": [{"axis": {"FromEnd": 1}, "partitions": 10, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 240, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 60, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 24}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 6, "column_partitions": 24, "inner_partitions": 10, "result_row_partitions": 10, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 24, "axes": [{"axis": {"FromEnd": 1}, "partitions": 10, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 240, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 60, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 24}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 4, "column_partitions": 72, "inner_partitions": 5, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 360, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 4, "column_partitions": 72, "inner_partitions": 5, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 360, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 1152, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 77, "column_partitions": 1, "inner_partitions": 19, "result_row_partitions": 1, "result_column_partitions": 19, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1463, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 19, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 77, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 19, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 19, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1463, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 19, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 77, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 19}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 1152, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 77, "column_partitions": 1, "inner_partitions": 19, "result_row_partitions": 1, "result_column_partitions": 19, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1463, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 19, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 77, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 19, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 19, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1463, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 19, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 77, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 19}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 128, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 23, "column_partitions": 9, "inner_partitions": 7, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1449, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 7, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 23, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 63, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 7, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 207, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 23, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 128, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 23, "column_partitions": 9, "inner_partitions": 7, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1449, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 7, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 23, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 63, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 7, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 207, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 23, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 72, "inner_partitions": 10, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 10, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 72, "inner_partitions": 10, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 10, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 288, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 19, "column_partitions": 4, "inner_partitions": 19, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1444, "replicas": 4, "axes": [{"axis": {"FromEnd": 1}, "partitions": 19, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 19, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 76, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 19, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 76, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 19, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 4}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 288, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 19, "column_partitions": 4, "inner_partitions": 19, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1444, "replicas": 4, "axes": [{"axis": {"FromEnd": 1}, "partitions": 19, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 19, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 76, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 19, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 76, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 19, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 4}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}], "1": [{"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1458, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1472, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1472, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1472, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 729, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1024, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1024, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1024, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 512}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 512, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 512}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 256}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 256, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 256}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 128}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 128, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 128}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 64}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 64, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 64}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 32, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 16}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 16, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 16}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 8}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 8, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 8}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}], "2": [{"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 2, "axes": [{"axis": {"FromStart": 1}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1472, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1472, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1472, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromStart": 1}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1024, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1024, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1024, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 512}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromStart": 1}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 512}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 256}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromStart": 1}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 256}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 128}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromStart": 1}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 128}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 64}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromStart": 1}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 64}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromStart": 1}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 16}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromStart": 1}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 16}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 8}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromStart": 1}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 8}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}], "3": [{"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1458, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1458, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1472, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1472, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1472, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1472, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 729, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 729, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1024, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1024, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1024, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1024, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 512}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 512, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 512, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 512}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 256}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 256, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 256, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 256}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 128}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 128, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 128, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 128}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 64}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 64, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 64, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 64}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 32, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 32, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 16}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 16, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 16, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 16}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 8}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 8, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 8, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 8}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}], "4": [{"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 216, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 216, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 432, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 216}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 216, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 216, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 432, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 216}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 216, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 216, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 432, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 216}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 216, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 216, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 432, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 216}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 216, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 216, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 432, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 216}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 216, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 216, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 432, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 216}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 216, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 216, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 432, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 216}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 216, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 216, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 432, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 216}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 216, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 216, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 648, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 216}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 216, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 216, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 648, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 216}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 216, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 216, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 648, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 216}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 216, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 216, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 648, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 216}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 6, "column_partitions": 122, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1464, "replicas": 122, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 244, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1464, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 122}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 6, "column_partitions": 122, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1464, "replicas": 122, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 244, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1464, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 122}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 6, "column_partitions": 122, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1464, "replicas": 122, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 244, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1464, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 122}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 6, "column_partitions": 122, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1464, "replicas": 122, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 244, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1464, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 122}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 4, "column_partitions": 122, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1464, "replicas": 122, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 366, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1464, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 122}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 4, "column_partitions": 122, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1464, "replicas": 122, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 366, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1464, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 122}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 73, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 73, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 146, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 73}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 73, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 73, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 146, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 73}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 73, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 73, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 146, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 73}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 73, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 73, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 146, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 73}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 73, "inner_partitions": 4, "result_row_partitions": 4, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 73, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 292, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 73}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 73, "inner_partitions": 4, "result_row_partitions": 4, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 73, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 292, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 73}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 216, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 216, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 648, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 216}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 216, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 216, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 648, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 216}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 6, "column_partitions": 122, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1464, "replicas": 122, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 244, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1464, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 122}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 6, "column_partitions": 122, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1464, "replicas": 122, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 244, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1464, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 122}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 6, "column_partitions": 122, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1464, "replicas": 122, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 244, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1464, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 122}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 6, "column_partitions": 122, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1464, "replicas": 122, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 244, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1464, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 122}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 4, "column_partitions": 122, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1464, "replicas": 122, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 366, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1464, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 122}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 4, "column_partitions": 122, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1464, "replicas": 122, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 366, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1464, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 122}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 216, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 216, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 432, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 648, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 216}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 216, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 216, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 432, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 648, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 216}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 216, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 216, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 432, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 648, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 216}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 216, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 216, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 432, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 648, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 216}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 216, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 216, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 432, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 648, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 3}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 216, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 216, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 432, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 648, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 3}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 216, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 216, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 432, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 648, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 3}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 216, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 216, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 432, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 648, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 3}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 216, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 216, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 648, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 216}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 216, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 216, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 648, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 216}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 9, "column_partitions": 54, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 54, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 162, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 27, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 54}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 9, "column_partitions": 54, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 54, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 162, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 27, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 54}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 9, "column_partitions": 54, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 54, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 162, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 27, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 54}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 9, "column_partitions": 54, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 54, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 162, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 27, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 54}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 73, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 73, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 146, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 146, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 146}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 73, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 73, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 146, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 146, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 146}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 73, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 2, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 73, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 146, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 146, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 146}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 73, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 2, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 73, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 146, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 146, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 146}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 122, "inner_partitions": 4, "result_row_partitions": 4, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1464, "replicas": 122, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 488, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1464, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 122}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 122, "inner_partitions": 4, "result_row_partitions": 4, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1464, "replicas": 122, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 488, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1464, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 122}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 73, "inner_partitions": 4, "result_row_partitions": 4, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 73, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 292, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 73}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 73, "inner_partitions": 4, "result_row_partitions": 4, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 73, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 292, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 73}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 216, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 216, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 432, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 648, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 216}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 216, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 216, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 432, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 648, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 216}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 216, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 216, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 432, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 648, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 216}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 216, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 216, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 432, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 648, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 216}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 216, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 216, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 432, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 648, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 3}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 216, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 216, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 432, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 648, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 3}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 216, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 216, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 432, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 648, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 3}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 216, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 216, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 432, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 648, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 3}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 73, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 73, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 146, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 73}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 73, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 73, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 146, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 73}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 73, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 73, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 146, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 73}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 73, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 73, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 146, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 73}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 4, "column_partitions": 122, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1464, "replicas": 122, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 366, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1464, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 122}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 4, "column_partitions": 122, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1464, "replicas": 122, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 366, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1464, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 122}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 73, "inner_partitions": 4, "result_row_partitions": 2, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 73, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 292, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 146, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 146}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 73, "inner_partitions": 4, "result_row_partitions": 2, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 73, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 292, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 146, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 146}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 9, "column_partitions": 54, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 54, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 162, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 27, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 54}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 9, "column_partitions": 54, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 54, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 162, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 27, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 54}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 216, "inner_partitions": 6, "result_row_partitions": 6, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 216, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 216}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 216, "inner_partitions": 6, "result_row_partitions": 6, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 216, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 216}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 4, "column_partitions": 73, "inner_partitions": 5, "result_row_partitions": 5, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 73, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 365, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 73}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 4, "column_partitions": 73, "inner_partitions": 5, "result_row_partitions": 5, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 73, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 365, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 73}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 9, "column_partitions": 54, "inner_partitions": 3, "result_row_partitions": 1, "result_column_partitions": 3, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 54, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 162, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 162, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 162}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 9, "column_partitions": 54, "inner_partitions": 3, "result_row_partitions": 1, "result_column_partitions": 3, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 54, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 162, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 162, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 162}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 96, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 36, "inner_partitions": 4, "result_row_partitions": 1, "result_column_partitions": 4, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 144}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 96, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 36, "inner_partitions": 4, "result_row_partitions": 1, "result_column_partitions": 4, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 144}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 4, "column_partitions": 61, "inner_partitions": 6, "result_row_partitions": 3, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1464, "replicas": 61, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 366, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 61, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1464, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 122}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 4, "column_partitions": 61, "inner_partitions": 6, "result_row_partitions": 3, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1464, "replicas": 61, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 366, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 61, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1464, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 122}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 54, "inner_partitions": 9, "result_row_partitions": 9, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 54, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 486, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 27, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 54}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 54, "inner_partitions": 9, "result_row_partitions": 9, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 54, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 486, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 27, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 54}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 96, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 36, "inner_partitions": 8, "result_row_partitions": 4, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 96, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 36, "inner_partitions": 8, "result_row_partitions": 4, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 144, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 12, "column_partitions": 24, "inner_partitions": 5, "result_row_partitions": 1, "result_column_partitions": 5, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 24, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 120, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 120, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 120}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 144, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 12, "column_partitions": 24, "inner_partitions": 5, "result_row_partitions": 1, "result_column_partitions": 5, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 24, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 120, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 120, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 120}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 216, "inner_partitions": 3, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 216, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 648, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 432, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 216}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 216, "inner_partitions": 3, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 216, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 648, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 432, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 216}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 96, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 36, "inner_partitions": 8, "result_row_partitions": 8, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 40, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 36}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 96, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 36, "inner_partitions": 8, "result_row_partitions": 8, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 40, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 36}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 128, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 9, "column_partitions": 27, "inner_partitions": 6, "result_row_partitions": 1, "result_column_partitions": 6, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 27, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 162, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 27, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 162, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 162}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 128, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 9, "column_partitions": 27, "inner_partitions": 6, "result_row_partitions": 1, "result_column_partitions": 6, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 27, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 162, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 27, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 162, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 162}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 192, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 9, "column_partitions": 18, "inner_partitions": 9, "result_row_partitions": 3, "result_column_partitions": 3, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 162, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 27, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 54}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 192, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 9, "column_partitions": 18, "inner_partitions": 9, "result_row_partitions": 3, "result_column_partitions": 3, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 162, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 27, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 54}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 122, "inner_partitions": 12, "result_row_partitions": 12, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1464, "replicas": 122, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 1464, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1464, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 122}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 122, "inner_partitions": 12, "result_row_partitions": 12, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1464, "replicas": 122, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 1464, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1464, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 122}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 144, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 24, "inner_partitions": 12, "result_row_partitions": 4, "result_column_partitions": 3, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 24, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 144, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 24, "inner_partitions": 12, "result_row_partitions": 4, "result_column_partitions": 3, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 24, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 256, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 13, "column_partitions": 14, "inner_partitions": 8, "result_row_partitions": 1, "result_column_partitions": 8, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1456, "replicas": 14, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 13, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 112, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 14, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1456, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 112, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 13, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 112}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 256, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 13, "column_partitions": 14, "inner_partitions": 8, "result_row_partitions": 1, "result_column_partitions": 8, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1456, "replicas": 14, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 13, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 112, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 14, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1456, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 112, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 13, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 112}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 192, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 9, "column_partitions": 18, "inner_partitions": 9, "result_row_partitions": 1, "result_column_partitions": 9, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 162, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 162, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 162}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 192, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 9, "column_partitions": 18, "inner_partitions": 9, "result_row_partitions": 1, "result_column_partitions": 9, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 162, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 162, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 162}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 73, "inner_partitions": 4, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 73, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 292, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 365, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 73}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 73, "inner_partitions": 4, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 73, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 292, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 365, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 73}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 320, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 11, "column_partitions": 11, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 12, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1452, "replicas": 11, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 11, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 132, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 11, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1452, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 132, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 11, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 132}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 320, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 11, "column_partitions": 11, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 12, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1452, "replicas": 11, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 11, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 132, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 11, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1452, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 132, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 11, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 132}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 128, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 27, "inner_partitions": 18, "result_row_partitions": 9, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 27, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 486, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 27, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 27, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 54}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 128, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 27, "inner_partitions": 18, "result_row_partitions": 9, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 27, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 486, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 27, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 27, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 54}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 128, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 27, "inner_partitions": 18, "result_row_partitions": 18, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 27, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 486, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 27, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 27, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 54, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 27}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 128, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 27, "inner_partitions": 18, "result_row_partitions": 18, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 27, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 486, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 27, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 27, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 54, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 27}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 432, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 23, "column_partitions": 8, "inner_partitions": 8, "result_row_partitions": 8, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 8, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 23, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 184, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 8}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 432, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 23, "column_partitions": 8, "inner_partitions": 8, "result_row_partitions": 8, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 8, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 23, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 184, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 8}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 4, "column_partitions": 73, "inner_partitions": 5, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 73, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 365, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 292, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 73}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 4, "column_partitions": 73, "inner_partitions": 5, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 73, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 365, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 292, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 73}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 1152, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 27, "column_partitions": 3, "inner_partitions": 18, "result_row_partitions": 1, "result_column_partitions": 18, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 3, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 27, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 54, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 27, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 54}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 1152, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 27, "column_partitions": 3, "inner_partitions": 18, "result_row_partitions": 1, "result_column_partitions": 18, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 3, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 27, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 54, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 27, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 54}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 128, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 9, "column_partitions": 27, "inner_partitions": 6, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 27, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 162, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 27, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 243, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 27, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 27}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 128, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 9, "column_partitions": 27, "inner_partitions": 6, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 27, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 162, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 27, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 243, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 27, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 27}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 96, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 36, "inner_partitions": 8, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 180, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 36}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 96, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 36, "inner_partitions": 8, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 180, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 36}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 54, "inner_partitions": 9, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 54, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 486, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 162, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 54}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 54, "inner_partitions": 9, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 54, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 486, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 162, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 54}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 144, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 24, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 24, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 120, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 24}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 144, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 24, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 24, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 120, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 24}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 128, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 27, "inner_partitions": 18, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 27, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 486, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 27, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 81, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 27, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 27}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 128, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 27, "inner_partitions": 18, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 27, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 486, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 27, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 81, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 27, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 27}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}], "5": [{"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1458, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1472, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1472, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1472, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 729, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1024, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1024, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1024, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 512}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 512, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 512}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 256}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 256, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 256}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 128}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 128, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 128}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 64}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 64, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 64}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 32, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}], "6": [{"operator": {"Slice": {"axis": 2, "start": 0, "length": 1152}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"Slice": {"axis": 2, "start": 0, "length": 1152}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}], "7": [{"operator": {"Slice": {"axis": 2, "start": 1152, "length": 1152}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"Slice": {"axis": 2, "start": 1152, "length": 1152}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}], "8": [{"operator": {"Slice": {"axis": 2, "start": 2304, "length": 1152}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"Slice": {"axis": 2, "start": 2304, "length": 1152}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}], "9": [{"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 16, "reversed": false}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": {"source_input": 0, "transform": {"split_axis": 2, "merge_axis": 0, "factor": 16, "reversed": false}, "unfused_cycles": 0, "unfused_exchange_cycles": 0}}], "10": [{"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 16, "reversed": false}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": {"source_input": 0, "transform": {"split_axis": 2, "merge_axis": 0, "factor": 16, "reversed": false}, "unfused_cycles": 0, "unfused_exchange_cycles": 0}}], "11": [{"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 16, "reversed": false}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": {"source_input": 0, "transform": {"split_axis": 2, "merge_axis": 0, "factor": 16, "reversed": false}, "unfused_cycles": 0, "unfused_exchange_cycles": 0}}], "12": [{"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 2, "columns": 4, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 2, "columns": 4, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 2, "columns": 2, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 2, "columns": 2, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 3, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 3, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 4, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 4, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 5, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 5, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 5, "columns": 4, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 5, "columns": 4, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 5, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 5, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 5, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 5, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 6, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 6, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 6, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 6, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 6, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 6, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 7, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 7, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 7, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 7, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 7, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 7, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 8, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 8, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 8, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 8, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 9, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 9, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 9, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 9, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 10, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 10, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 10, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 10, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 11, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 11, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 11, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 11, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 12, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 12, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 13, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 13, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 14, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 14, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 15, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 15, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 16, "columns": 2, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 16, "columns": 2, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 16, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 16, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 2, "columns": 4, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 2, "columns": 4, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 2, "columns": 2, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 2, "columns": 2, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 3, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 3, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 4, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 4, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 5, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 5, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 5, "columns": 4, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 5, "columns": 4, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 5, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 5, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 5, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 5, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 6, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 6, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 6, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 6, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 6, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 6, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 7, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 7, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 7, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 7, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 7, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 7, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 8, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 8, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 8, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 8, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 9, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 9, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 9, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 9, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 10, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 10, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 10, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 10, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 11, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 11, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 11, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 11, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 12, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 12, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 13, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 13, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 14, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 14, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 15, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 15, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 16, "columns": 2, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 16, "columns": 2, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 16, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 2, "columns": 23, "inner": 1}, "probability_value": {"rows": 16, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 2, "columns": 4, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 2, "columns": 4, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 2, "columns": 2, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 2, "columns": 2, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 3, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 3, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 4, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 4, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 5, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 5, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 5, "columns": 4, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 5, "columns": 4, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 5, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 5, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 5, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 5, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 6, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 6, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 6, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 6, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 6, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 6, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 7, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 7, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 7, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 7, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 7, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 7, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 8, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 8, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 8, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 8, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 9, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 9, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 9, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 9, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 10, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 10, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 10, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 10, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 11, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 11, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 11, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 11, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 12, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 12, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 13, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 13, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 14, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 14, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 15, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 15, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 16, "columns": 2, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 16, "columns": 2, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 16, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 3, "columns": 15, "inner": 1}, "probability_value": {"rows": 16, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 2, "columns": 4, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 2, "columns": 4, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 2, "columns": 2, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 2, "columns": 2, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 3, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 3, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 4, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 4, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 5, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 5, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 5, "columns": 4, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 5, "columns": 4, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 5, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 5, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 5, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 5, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 6, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 6, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 6, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 6, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 6, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 6, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 7, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 7, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 7, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 7, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 7, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 7, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 8, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 8, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 8, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 8, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 9, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 9, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 9, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 9, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 10, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 10, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 10, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 10, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 11, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 11, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 11, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 11, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 12, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 12, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 13, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 13, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 14, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 14, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 15, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 15, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 16, "columns": 2, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 16, "columns": 2, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 16, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 4, "columns": 11, "inner": 1}, "probability_value": {"rows": 16, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 2, "columns": 4, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 2, "columns": 4, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 2, "columns": 2, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 2, "columns": 2, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 3, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 3, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 4, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 4, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 5, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 5, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 5, "columns": 4, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 5, "columns": 4, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 5, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 5, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 5, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 5, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 6, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 6, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 6, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 6, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 6, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 6, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 7, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 7, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 7, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 7, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 7, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 7, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 8, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 8, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 8, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 8, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 9, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 9, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 9, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 9, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 10, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 10, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 10, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 10, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 11, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 11, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 11, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 11, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 12, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 12, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 13, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 13, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 14, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 14, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 15, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 15, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 16, "columns": 2, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 16, "columns": 2, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 16, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 5, "columns": 9, "inner": 1}, "probability_value": {"rows": 16, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 2, "columns": 4, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 2, "columns": 4, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 2, "columns": 2, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 2, "columns": 2, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 3, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 3, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 4, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 4, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 5, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 5, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 5, "columns": 4, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 5, "columns": 4, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 5, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 5, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 5, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 5, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 6, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 6, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 6, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 6, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 6, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 6, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 7, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 7, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 7, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 7, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 7, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 7, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 8, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 8, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 8, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 8, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 9, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 9, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 9, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 9, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 10, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 10, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 10, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 10, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 11, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 11, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 11, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 11, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 12, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 12, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 13, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 13, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 14, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 14, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 15, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 15, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 16, "columns": 2, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 16, "columns": 2, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 16, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 6, "columns": 7, "inner": 1}, "probability_value": {"rows": 16, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 2, "columns": 4, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 2, "columns": 4, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 2, "columns": 2, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 2, "columns": 2, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 3, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 3, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 4, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 4, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 5, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 5, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 5, "columns": 4, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 5, "columns": 4, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 5, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 5, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 5, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 5, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 6, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 6, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 6, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 6, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 6, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 6, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 7, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 7, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 7, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 7, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 7, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 7, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 8, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 8, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 8, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 8, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 9, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 9, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 9, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 9, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 10, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 10, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 10, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 10, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 11, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 11, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 11, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 11, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 12, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 12, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 13, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 13, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 14, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 14, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 15, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 15, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 16, "columns": 2, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 16, "columns": 2, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 16, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 7, "columns": 6, "inner": 1}, "probability_value": {"rows": 16, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 2, "columns": 4, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 2, "columns": 4, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 2, "columns": 2, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 2, "columns": 2, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 3, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 3, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 4, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 4, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 5, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 5, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 5, "columns": 4, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 5, "columns": 4, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 5, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 5, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 5, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 5, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 6, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 6, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 6, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 6, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 6, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 6, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 7, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 7, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 7, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 7, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 7, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 7, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 8, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 8, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 8, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 8, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 9, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 9, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 9, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 9, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 10, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 10, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 10, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 10, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 11, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 11, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 11, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 11, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 12, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 12, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 13, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 13, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 14, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 14, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 15, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 15, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 16, "columns": 2, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 16, "columns": 2, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 16, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 8, "columns": 5, "inner": 1}, "probability_value": {"rows": 16, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 2, "columns": 4, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 2, "columns": 4, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 2, "columns": 2, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 2, "columns": 2, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 3, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 3, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 4, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 4, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 5, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 5, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 5, "columns": 4, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 5, "columns": 4, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 5, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 5, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 5, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 5, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 6, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 6, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 6, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 6, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 6, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 6, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 7, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 7, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 7, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 7, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 7, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 7, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 8, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 8, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 8, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 8, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 9, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 9, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 9, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 9, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 10, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 10, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 10, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 10, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 11, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 11, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 11, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 11, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 12, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 12, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 13, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 13, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 14, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 14, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 15, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 15, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 16, "columns": 2, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 16, "columns": 2, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 16, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 9, "columns": 5, "inner": 1}, "probability_value": {"rows": 16, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 2, "columns": 4, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 2, "columns": 4, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 2, "columns": 2, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 2, "columns": 2, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 3, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 3, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 4, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 4, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 5, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 5, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 5, "columns": 4, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 5, "columns": 4, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 5, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 5, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 5, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 5, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 6, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 6, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 6, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 6, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 6, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 6, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 7, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 7, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 7, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 7, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 7, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 7, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 8, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 8, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 8, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 8, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 9, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 9, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 9, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 9, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 10, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 10, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 10, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 10, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 11, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 11, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 11, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 11, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 12, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 12, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 13, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 13, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 14, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 14, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 15, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 15, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 16, "columns": 2, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 16, "columns": 2, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 16, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 10, "columns": 4, "inner": 1}, "probability_value": {"rows": 16, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 2, "columns": 4, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 2, "columns": 4, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 2, "columns": 2, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 2, "columns": 2, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 3, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 3, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 4, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 4, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 5, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 5, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 5, "columns": 4, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 5, "columns": 4, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 5, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 5, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 5, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 5, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 6, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 6, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 6, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 6, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 6, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 6, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 7, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 7, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 7, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 7, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 7, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 7, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 8, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 8, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 8, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 8, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 9, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 9, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 9, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 9, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 10, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 10, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 10, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 10, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 11, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 11, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 11, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 11, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 12, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 12, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 13, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 13, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 14, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 14, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 15, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 15, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 16, "columns": 2, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 16, "columns": 2, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 16, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 11, "columns": 4, "inner": 1}, "probability_value": {"rows": 16, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 4, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 4, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 2, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 2, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 4, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 4, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 6, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 6, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 6, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 6, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 6, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 6, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 7, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 7, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 7, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 7, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 7, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 7, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 8, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 8, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 8, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 8, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 9, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 9, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 9, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 9, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 10, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 10, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 10, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 10, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 11, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 11, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 11, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 11, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 12, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 12, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 13, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 13, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 14, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 14, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 15, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 15, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 16, "columns": 2, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 16, "columns": 2, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 16, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 12, "columns": 3, "inner": 1}, "probability_value": {"rows": 16, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 4, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 4, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 2, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 2, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 4, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 4, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 6, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 6, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 6, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 6, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 6, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 6, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 7, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 7, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 7, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 7, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 7, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 7, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 8, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 8, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 8, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 8, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 9, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 9, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 9, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 9, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 10, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 10, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 10, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 10, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 11, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 11, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 11, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 11, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 12, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 12, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 13, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 13, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 14, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 14, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 15, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 15, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 16, "columns": 2, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 16, "columns": 2, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 16, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 13, "columns": 3, "inner": 1}, "probability_value": {"rows": 16, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 4, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 4, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 2, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 2, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 4, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 4, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 6, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 6, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 6, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 6, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 6, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 6, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 7, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 7, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 7, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 7, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 7, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 7, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 8, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 8, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 8, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 8, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 9, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 9, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 9, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 9, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 10, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 10, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 10, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 10, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 11, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 11, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 11, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 11, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 12, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 12, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 13, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 13, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 14, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 14, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 15, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 15, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 16, "columns": 2, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 16, "columns": 2, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 16, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 14, "columns": 3, "inner": 1}, "probability_value": {"rows": 16, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 4, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 4, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 2, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 2, "columns": 2, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 3, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 4, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 4, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 6, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 6, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 6, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 6, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 6, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 6, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 7, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 7, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 7, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 7, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 7, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 7, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 8, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 8, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 8, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 8, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 9, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 9, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 9, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 9, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 10, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 10, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 10, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 10, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 11, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 11, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 11, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 11, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 12, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 12, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 13, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 13, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 14, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 14, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 15, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 15, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 16, "columns": 2, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 16, "columns": 2, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 16, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 15, "columns": 3, "inner": 1}, "probability_value": {"rows": 16, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 2, "columns": 4, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 2, "columns": 4, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 2, "columns": 3, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 2, "columns": 2, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 2, "columns": 2, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 3, "columns": 3, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 3, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 3, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 4, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 4, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 4, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 5, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 5, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 5, "columns": 4, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 5, "columns": 4, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 5, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 5, "columns": 3, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 5, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 5, "columns": 2, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 6, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 6, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 6, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 6, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 6, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 6, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 7, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 7, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 7, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 7, "columns": 3, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 7, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 7, "columns": 2, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 8, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 8, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 8, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 8, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 9, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 9, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 9, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 9, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 10, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 10, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 10, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 10, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 11, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 11, "columns": 4, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 11, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 11, "columns": 2, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 12, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 12, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 13, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 13, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 14, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 14, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 15, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 15, "columns": 3, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 16, "columns": 2, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 16, "columns": 2, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 16, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, -4], "query_key": {"rows": 16, "columns": 2, "inner": 1}, "probability_value": {"rows": 16, "columns": 1, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}], "13": [{"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 16, "reversed": true}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 46, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}], "14": [{"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 72, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 72, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 72, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 72, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 72, "inner_partitions": 4, "result_row_partitions": 4, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 72, "inner_partitions": 4, "result_row_partitions": 4, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 6, "column_partitions": 72, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 216, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 6, "column_partitions": 72, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 216, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 72, "inner_partitions": 4, "result_row_partitions": 4, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 72, "inner_partitions": 4, "result_row_partitions": 4, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 72, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 72, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 72, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 72, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 6, "column_partitions": 72, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 216, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 6, "column_partitions": 72, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 216, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 4, "column_partitions": 72, "inner_partitions": 5, "result_row_partitions": 5, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 360, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 4, "column_partitions": 72, "inner_partitions": 5, "result_row_partitions": 5, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 360, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 72, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 72, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 72, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 72, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 72, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 10}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 72, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 10}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 72, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 10}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 72, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 10}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 36, "inner_partitions": 4, "result_row_partitions": 2, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 36, "inner_partitions": 4, "result_row_partitions": 2, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 4, "column_partitions": 72, "inner_partitions": 5, "result_row_partitions": 5, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 360, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 4, "column_partitions": 72, "inner_partitions": 5, "result_row_partitions": 5, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 360, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 36, "inner_partitions": 4, "result_row_partitions": 4, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 40, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 36}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 36, "inner_partitions": 4, "result_row_partitions": 4, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 40, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 36}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 6, "column_partitions": 72, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 216, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 6, "column_partitions": 72, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 216, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 13, "column_partitions": 37, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1443, "replicas": 37, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 13, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 111, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 37, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1443, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 37, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 39, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 37}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 13, "column_partitions": 37, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1443, "replicas": 37, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 13, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 111, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 37, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1443, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 37, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 39, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 37}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 6, "column_partitions": 72, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 216, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 6, "column_partitions": 72, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 216, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 72, "inner_partitions": 6, "result_row_partitions": 6, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 432, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 72, "inner_partitions": 6, "result_row_partitions": 6, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 432, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 72, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 72, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 72, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 72, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 72, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 10}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 72, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 10}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 72, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 10}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 72, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 10}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 72, "inner_partitions": 4, "result_row_partitions": 4, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 72, "inner_partitions": 4, "result_row_partitions": 4, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 36, "inner_partitions": 4, "result_row_partitions": 2, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 36, "inner_partitions": 4, "result_row_partitions": 2, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 36, "inner_partitions": 4, "result_row_partitions": 4, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 40, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 36}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 36, "inner_partitions": 4, "result_row_partitions": 4, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 40, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 36}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 72, "inner_partitions": 6, "result_row_partitions": 6, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 432, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 72, "inner_partitions": 6, "result_row_partitions": 6, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 432, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 8, "column_partitions": 36, "inner_partitions": 5, "result_row_partitions": 5, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 180, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 40, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 36}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 8, "column_partitions": 36, "inner_partitions": 5, "result_row_partitions": 5, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 180, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 40, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 36}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 72, "inner_partitions": 4, "result_row_partitions": 4, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 72, "inner_partitions": 4, "result_row_partitions": 4, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 36, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 36, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 36, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 2, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 36, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 2, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 36, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 40, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 36}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 36, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 40, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 36}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 36, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 40, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 36}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 36, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 40, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 36}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 13, "column_partitions": 37, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1443, "replicas": 37, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 13, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 111, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 37, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1443, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 37, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 39, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 37}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 13, "column_partitions": 37, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1443, "replicas": 37, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 13, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 111, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 37, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1443, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 37, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 39, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 37}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 24, "inner_partitions": 6, "result_row_partitions": 2, "result_column_partitions": 3, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 24, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 24, "inner_partitions": 6, "result_row_partitions": 2, "result_column_partitions": 3, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 24, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 24, "inner_partitions": 3, "result_row_partitions": 1, "result_column_partitions": 3, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 24, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 24, "inner_partitions": 3, "result_row_partitions": 1, "result_column_partitions": 3, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 24, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 36, "inner_partitions": 8, "result_row_partitions": 4, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 36, "inner_partitions": 8, "result_row_partitions": 4, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 36, "inner_partitions": 8, "result_row_partitions": 8, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 40, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 36}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 36, "inner_partitions": 8, "result_row_partitions": 8, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 40, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 36}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 18, "inner_partitions": 4, "result_row_partitions": 1, "result_column_partitions": 4, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 18, "inner_partitions": 4, "result_row_partitions": 1, "result_column_partitions": 4, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 72, "inner_partitions": 9, "result_row_partitions": 9, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 648, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 72, "inner_partitions": 9, "result_row_partitions": 9, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 648, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 6, "column_partitions": 72, "inner_partitions": 3, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 216, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 432, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 6, "column_partitions": 72, "inner_partitions": 3, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 216, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 432, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 96, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 12, "inner_partitions": 6, "result_row_partitions": 1, "result_column_partitions": 6, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 12, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 96, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 12, "inner_partitions": 6, "result_row_partitions": 1, "result_column_partitions": 6, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 12, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 24, "inner_partitions": 12, "result_row_partitions": 4, "result_column_partitions": 3, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 24, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 24, "inner_partitions": 12, "result_row_partitions": 4, "result_column_partitions": 3, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 24, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 24, "inner_partitions": 12, "result_row_partitions": 12, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 24, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 60, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 24}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 24, "inner_partitions": 12, "result_row_partitions": 12, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 24, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 60, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 24}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 128, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 18, "column_partitions": 9, "inner_partitions": 9, "result_row_partitions": 3, "result_column_partitions": 3, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 81, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 27, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 54, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 27}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 128, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 18, "column_partitions": 9, "inner_partitions": 9, "result_row_partitions": 3, "result_column_partitions": 3, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 81, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 27, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 54, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 27}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 128, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 9, "inner_partitions": 8, "result_row_partitions": 1, "result_column_partitions": 8, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 128, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 9, "inner_partitions": 8, "result_row_partitions": 1, "result_column_partitions": 8, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 72, "inner_partitions": 4, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 360, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 72, "inner_partitions": 4, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 360, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 96, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 24, "column_partitions": 12, "inner_partitions": 5, "result_row_partitions": 1, "result_column_partitions": 5, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 12, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 24, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 60, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 60, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 24, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 60}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 96, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 24, "column_partitions": 12, "inner_partitions": 5, "result_row_partitions": 1, "result_column_partitions": 5, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 12, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 24, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 60, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 60, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 24, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 60}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 18, "result_row_partitions": 18, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 18, "result_row_partitions": 18, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 128, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 9, "column_partitions": 9, "inner_partitions": 18, "result_row_partitions": 9, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 162, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 81, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 128, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 9, "column_partitions": 9, "inner_partitions": 18, "result_row_partitions": 9, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 162, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 81, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 192, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 27, "column_partitions": 6, "inner_partitions": 9, "result_row_partitions": 1, "result_column_partitions": 9, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 27, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 54, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 27, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 54}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 192, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 27, "column_partitions": 6, "inner_partitions": 9, "result_row_partitions": 1, "result_column_partitions": 9, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 27, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 54, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 27, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 54}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 4, "column_partitions": 72, "inner_partitions": 5, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 360, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 4, "column_partitions": 72, "inner_partitions": 5, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 360, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 96, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 12, "inner_partitions": 12, "result_row_partitions": 2, "result_column_partitions": 6, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 12, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 96, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 12, "inner_partitions": 12, "result_row_partitions": 2, "result_column_partitions": 6, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 12, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 576, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 61, "column_partitions": 2, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 12, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1464, "replicas": 2, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 61, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 24, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1464, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 61, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 24}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 576, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 61, "column_partitions": 2, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 12, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1464, "replicas": 2, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 61, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 24, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1464, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 61, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 24}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 384, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 27, "column_partitions": 3, "inner_partitions": 18, "result_row_partitions": 1, "result_column_partitions": 18, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 3, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 27, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 54, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 27, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 54}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 384, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 27, "column_partitions": 3, "inner_partitions": 18, "result_row_partitions": 1, "result_column_partitions": 18, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 3, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 27, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 54, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 27, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 54}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 144, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 8, "inner_partitions": 36, "result_row_partitions": 4, "result_column_partitions": 9, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 8, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 144, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 8, "inner_partitions": 36, "result_row_partitions": 4, "result_column_partitions": 9, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 8, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 24, "inner_partitions": 6, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 24, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 240, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 24}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 24, "inner_partitions": 6, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 24, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 240, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 24}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 144, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 8, "inner_partitions": 36, "result_row_partitions": 36, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 8, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 180, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 8}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 144, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 8, "inner_partitions": 36, "result_row_partitions": 36, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 8, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 180, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 8}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 36, "inner_partitions": 8, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 180, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 36}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 36, "inner_partitions": 8, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 180, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 36}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 9, "column_partitions": 18, "inner_partitions": 9, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 162, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 162, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 9, "column_partitions": 18, "inner_partitions": 9, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 162, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 162, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 24, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 24, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 120, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 24}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 24, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 24, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 120, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 24}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 128, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 9, "column_partitions": 9, "inner_partitions": 18, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 162, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 81, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 128, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 9, "column_partitions": 9, "inner_partitions": 18, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 162, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 81, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 144, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 8, "inner_partitions": 36, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 8, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 40, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 8}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 144, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 8, "inner_partitions": 36, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 8, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 40, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 8}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}], "15": [{"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1458, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1472, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1472, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1472, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 729, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1024, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1024, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1024, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 512}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 512, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 512}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 256}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 256, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 256}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 128}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 128, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 128}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 64}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 64, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 64}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 32, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 16}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 16, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 16}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 8}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 8, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 8}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}], "16": [{"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1472, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1472, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1472, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1024, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1024, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1024, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 512}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 512}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 512}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 256}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 256}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 256}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 128}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 128}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 128}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 64}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 64}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 64}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 16}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 16}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 16}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 8}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 8}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 8}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}], "17": [{"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1458, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1458, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1472, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1472, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1472, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1472, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 729, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 729, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1024, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1024, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1024, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1024, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 512}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 512, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 512, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 512}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 256}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 256, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 256, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 256}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 128}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 128, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 128, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 128}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 64}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 64, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 64, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 64}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 32, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 32, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 16}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 16, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 16, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 16}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 8}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 8, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 8, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 8}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}], "18": [{"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 147, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 147, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 294, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 147, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 147, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 147}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 147, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 147, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 294, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 147, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 147, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 147}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 147, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 147, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 294, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 147, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 147, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 147}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 147, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 147, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 294, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 147, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 147, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 147}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 8, "column_partitions": 92, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 92, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 184, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 92}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 8, "column_partitions": 92, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 92, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 184, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 92}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 8, "column_partitions": 92, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 92, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 184, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 92}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 8, "column_partitions": 92, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 92, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 184, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 92}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 147, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 147, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 294, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 147, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 147, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 147}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 147, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 147, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 294, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 147, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 147, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 147}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 147, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 147, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 294, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 147, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 147, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 147}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 147, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 147, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 294, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 147, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 147, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 147}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 4, "column_partitions": 92, "inner_partitions": 4, "result_row_partitions": 4, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 92, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 368, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 92}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 4, "column_partitions": 92, "inner_partitions": 4, "result_row_partitions": 4, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 92, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 368, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 92}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 8, "column_partitions": 92, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 92, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 184, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 184}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 8, "column_partitions": 92, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 92, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 184, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 184}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 8, "column_partitions": 92, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 2, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 92, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 184, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 184}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 8, "column_partitions": 92, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 2, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 92, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 184, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 184}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 98, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 98, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 294, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 98, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 98, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 15, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 98}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 98, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 98, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 294, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 98, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 98, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 15, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 98}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 8, "column_partitions": 92, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 92, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 184, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 92}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 8, "column_partitions": 92, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 92, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 184, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 92}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 8, "column_partitions": 92, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 92, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 184, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 92}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 8, "column_partitions": 92, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 92, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 184, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 92}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 7, "column_partitions": 70, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 70, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 7, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 210, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 70, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 70, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 70}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 7, "column_partitions": 70, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 70, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 7, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 210, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 70, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 70, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 70}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 4, "column_partitions": 92, "inner_partitions": 4, "result_row_partitions": 4, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 92, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 368, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 92}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 4, "column_partitions": 92, "inner_partitions": 4, "result_row_partitions": 4, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 92, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 368, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 92}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 98, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 98, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 294, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 98, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 98, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 15, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 98}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 98, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 98, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 294, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 98, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 98, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 15, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 98}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 80, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 9, "column_partitions": 54, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 54, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 162, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 27, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 54}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 80, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 9, "column_partitions": 54, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 54, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 162, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 27, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 54}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 8, "column_partitions": 92, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 92, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 184, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 184}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 8, "column_partitions": 92, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 92, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 184, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 184}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 8, "column_partitions": 92, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 2, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 92, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 184, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 184}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 8, "column_partitions": 92, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 2, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 92, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 184, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 184}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 73, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 73, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 146, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 146, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 146}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 73, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 73, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 146, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 146, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 146}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 73, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 2, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 73, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 146, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 146, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 146}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 73, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 2, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 73, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 146, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 146, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 146}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 147, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 147, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 294, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 147, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 735, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 147, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 147}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 147, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 147, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 294, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 147, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 735, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 147, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 147}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 147, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 147, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 294, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 147, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 735, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 147, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 147}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 147, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 147, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 294, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 147, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 735, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 147, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 147}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 147, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 147, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 294, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 147, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 735, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 147, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 5}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 147, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 147, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 294, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 147, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 735, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 147, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 5}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 147, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 147, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 294, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 147, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 735, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 147, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 5}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 147, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 147, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 294, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 147, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 735, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 147, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 5}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 269, "inner_partitions": 5, "result_row_partitions": 5, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1345, "replicas": 269, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 1345, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 269, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1345, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 269, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 269}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 269, "inner_partitions": 5, "result_row_partitions": 5, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1345, "replicas": 269, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 1345, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 269, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1345, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 269, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 269}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 73, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 73, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 146, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 73}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 73, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 73, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 146, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 73}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 73, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 73, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 146, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 73}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 73, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 73, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 146, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 73}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 269, "inner_partitions": 5, "result_row_partitions": 5, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1345, "replicas": 269, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 1345, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 269, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1345, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 269, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 269}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 269, "inner_partitions": 5, "result_row_partitions": 5, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1345, "replicas": 269, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 1345, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 269, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1345, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 269, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 269}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 73, "inner_partitions": 4, "result_row_partitions": 2, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 73, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 292, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 146, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 146}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 73, "inner_partitions": 4, "result_row_partitions": 2, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 73, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 292, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 146, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 146}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 7, "column_partitions": 70, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 70, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 7, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 210, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 70, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 70, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 70}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 7, "column_partitions": 70, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 70, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 7, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 210, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 70, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 70, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 70}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 98, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 98, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 294, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 98, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 98, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 15, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 98}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 98, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 98, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 294, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 98, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 98, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 15, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 98}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 73, "inner_partitions": 4, "result_row_partitions": 4, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 73, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 292, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 73}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 73, "inner_partitions": 4, "result_row_partitions": 4, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 73, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 292, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 73}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 80, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 9, "column_partitions": 54, "inner_partitions": 3, "result_row_partitions": 1, "result_column_partitions": 3, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 54, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 162, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 162, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 162}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 80, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 9, "column_partitions": 54, "inner_partitions": 3, "result_row_partitions": 1, "result_column_partitions": 3, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 54, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 162, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 162, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 162}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 4, "column_partitions": 92, "inner_partitions": 4, "result_row_partitions": 2, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 92, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 368, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 184}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 4, "column_partitions": 92, "inner_partitions": 4, "result_row_partitions": 2, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 92, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 368, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 184}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 8, "column_partitions": 92, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 92, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 184, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 736, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 92}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 8, "column_partitions": 92, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 92, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 184, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 736, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 92}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 8, "column_partitions": 92, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 92, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 184, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 736, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 92}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 8, "column_partitions": 92, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 92, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 184, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 736, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 92}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 8, "column_partitions": 92, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 92, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 184, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 736, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 8}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 8, "column_partitions": 92, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 92, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 184, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 736, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 8}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 8, "column_partitions": 92, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 92, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 184, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 736, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 8}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 8, "column_partitions": 92, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 92, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 184, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 736, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 8}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 96, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 8, "column_partitions": 46, "inner_partitions": 4, "result_row_partitions": 2, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 46, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 184, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 46, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 92}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 96, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 8, "column_partitions": 46, "inner_partitions": 4, "result_row_partitions": 2, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 46, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 184, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 46, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 92}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 7, "column_partitions": 70, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 70, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 7, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 210, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 70, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 70, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 70}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 7, "column_partitions": 70, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 70, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 7, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 210, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 70, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 70, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 70}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 147, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 147, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 294, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 147, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 735, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 147, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 147}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 147, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 147, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 294, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 147, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 735, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 147, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 147}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 147, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 147, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 294, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 147, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 735, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 147, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 147}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 147, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 147, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 294, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 147, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 735, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 147, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 147}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 96, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 8, "column_partitions": 46, "inner_partitions": 4, "result_row_partitions": 1, "result_column_partitions": 4, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 46, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 184, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 46, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 184}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 96, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 8, "column_partitions": 46, "inner_partitions": 4, "result_row_partitions": 1, "result_column_partitions": 4, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 46, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 184, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 46, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 184}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 96, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 49, "inner_partitions": 6, "result_row_partitions": 2, "result_column_partitions": 3, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 49, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 294, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 49, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 147, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 147}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 96, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 49, "inner_partitions": 6, "result_row_partitions": 2, "result_column_partitions": 3, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 49, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 294, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 49, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 147, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 147}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 128, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 7, "column_partitions": 35, "inner_partitions": 6, "result_row_partitions": 6, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 35, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 7, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 210, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 35, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 35, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 42, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 35}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 128, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 7, "column_partitions": 35, "inner_partitions": 6, "result_row_partitions": 6, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 35, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 7, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 210, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 35, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 35, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 42, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 35}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 92, "inner_partitions": 8, "result_row_partitions": 8, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 92, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 736, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 92}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 92, "inner_partitions": 8, "result_row_partitions": 8, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 92, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 736, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 92}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 80, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 54, "inner_partitions": 9, "result_row_partitions": 9, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 54, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 486, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 27, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 54}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 80, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 54, "inner_partitions": 9, "result_row_partitions": 9, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 54, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 486, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 27, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 54}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 160, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 9, "column_partitions": 27, "inner_partitions": 6, "result_row_partitions": 1, "result_column_partitions": 6, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 27, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 162, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 27, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 162, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 162}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 160, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 9, "column_partitions": 27, "inner_partitions": 6, "result_row_partitions": 1, "result_column_partitions": 6, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 27, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 162, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 27, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 162, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 162}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 96, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 4, "column_partitions": 46, "inner_partitions": 8, "result_row_partitions": 4, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 46, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 368, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 46, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 92}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 96, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 4, "column_partitions": 46, "inner_partitions": 8, "result_row_partitions": 4, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 46, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 368, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 46, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 92}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 160, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 29, "inner_partitions": 5, "result_row_partitions": 1, "result_column_partitions": 5, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1450, "replicas": 29, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 145, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 29, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1450, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 145, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 145}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 160, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 29, "inner_partitions": 5, "result_row_partitions": 1, "result_column_partitions": 5, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1450, "replicas": 29, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 145, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 29, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1450, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 145, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 145}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 98, "inner_partitions": 3, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 98, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 294, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 98, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 490, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 98, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 98}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 98, "inner_partitions": 3, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 98, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 294, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 98, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 490, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 98, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 98}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 80, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 54, "inner_partitions": 9, "result_row_partitions": 3, "result_column_partitions": 3, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 54, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 486, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 162, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 162}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 80, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 54, "inner_partitions": 9, "result_row_partitions": 3, "result_column_partitions": 3, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 54, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 486, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 162, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 162}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 240, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 18, "inner_partitions": 8, "result_row_partitions": 1, "result_column_partitions": 8, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 144}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 240, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 18, "inner_partitions": 8, "result_row_partitions": 1, "result_column_partitions": 8, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 144}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 144, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 4, "column_partitions": 30, "inner_partitions": 12, "result_row_partitions": 4, "result_column_partitions": 3, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 30, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 360, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 30, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 90, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 90}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 144, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 4, "column_partitions": 30, "inner_partitions": 12, "result_row_partitions": 4, "result_column_partitions": 3, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 30, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 360, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 30, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 90, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 90}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 240, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 9, "column_partitions": 18, "inner_partitions": 9, "result_row_partitions": 1, "result_column_partitions": 9, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 162, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 162, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 162}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 240, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 9, "column_partitions": 18, "inner_partitions": 9, "result_row_partitions": 1, "result_column_partitions": 9, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 162, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 162, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 162}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 144, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 4, "column_partitions": 30, "inner_partitions": 12, "result_row_partitions": 12, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 30, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 360, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 30, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 30, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 48, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 30}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 144, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 4, "column_partitions": 30, "inner_partitions": 12, "result_row_partitions": 12, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 30, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 360, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 30, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 30, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 48, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 30}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 4, "column_partitions": 92, "inner_partitions": 4, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 92, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 368, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 368, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 92}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 4, "column_partitions": 92, "inner_partitions": 4, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 92, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 368, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 368, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 92}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 288, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 8, "column_partitions": 15, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 12, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 15, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 180, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 15, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 180, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 180}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 288, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 8, "column_partitions": 15, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 12, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 15, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 180, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 15, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 180, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 180}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 544, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 23, "column_partitions": 8, "inner_partitions": 8, "result_row_partitions": 4, "result_column_partitions": 2, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 8, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 23, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 16, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 92, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 16}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 544, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 23, "column_partitions": 8, "inner_partitions": 8, "result_row_partitions": 4, "result_column_partitions": 2, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 8, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 23, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 16, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 92, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 16}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 269, "inner_partitions": 5, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1345, "replicas": 269, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 1345, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 269, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 269, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 269, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 269}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 269, "inner_partitions": 5, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1345, "replicas": 269, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 1345, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 269, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 269, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 269, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 269}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 160, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 9, "column_partitions": 27, "inner_partitions": 6, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 27, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 162, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 27, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 243, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 27, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 27}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 160, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 9, "column_partitions": 27, "inner_partitions": 6, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 27, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 162, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 27, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 243, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 27, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 27}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 92, "inner_partitions": 8, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 92, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 736, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 184, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 92}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 92, "inner_partitions": 8, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 92, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 736, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 184, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 92, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 92}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 80, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 54, "inner_partitions": 9, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 54, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 486, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 162, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 54}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 80, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 54, "inner_partitions": 9, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 54, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 486, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 162, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 54}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 144, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 4, "column_partitions": 30, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 30, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 360, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 30, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 120, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 30, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 30}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 144, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 4, "column_partitions": 30, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 30, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 360, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 30, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 120, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 30, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 30}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}], "19": [{"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1458, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1472, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1472, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1472, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 729, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1024, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1024, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1024, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 512}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 512, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 512}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 256}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 256, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 256}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 128}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 128, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 128}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 64}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 64, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 64}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 32, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}], "20": [{"operator": "Gelu", "dispatch": {"Pointwise": {"kernel": "Gelu"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 32, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 32, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Gelu", "dispatch": {"Pointwise": {"kernel": "Gelu"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 64, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 64, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Gelu", "dispatch": {"Pointwise": {"kernel": "Gelu"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 128, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 128, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Gelu", "dispatch": {"Pointwise": {"kernel": "Gelu"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 256, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 256, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Gelu", "dispatch": {"Pointwise": {"kernel": "Gelu"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 512, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 512, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Gelu", "dispatch": {"Pointwise": {"kernel": "Gelu"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 1024, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 1024, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Gelu", "dispatch": {"Pointwise": {"kernel": "Gelu"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 1472, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 1472, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Gelu", "dispatch": {"Pointwise": {"kernel": "Gelu"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}], "21": [{"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 864, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 4, "column_partitions": 72, "inner_partitions": 5, "result_row_partitions": 5, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 864, "padding_multiple": 864, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 864, "column_block": 16}}}, "tiling": {"tile_count": 360, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 864, "padding_multiple": 864, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 864, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 4, "column_partitions": 72, "inner_partitions": 5, "result_row_partitions": 5, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 864, "padding_multiple": 864, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 864, "column_block": 16}}}, "tiling": {"tile_count": 360, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 864, "padding_multiple": 864, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 1088, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 72, "inner_partitions": 4, "result_row_partitions": 4, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 1088, "padding_multiple": 1088, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 1088, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 1088, "padding_multiple": 1088, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 1088, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 72, "inner_partitions": 4, "result_row_partitions": 4, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 1088, "padding_multiple": 1088, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 1088, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 1088, "padding_multiple": 1088, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 448, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 72, "inner_partitions": 10, "result_row_partitions": 10, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 10, "padding_groups": 1, "block_size": 448, "padding_multiple": 448, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 448, "column_block": 16}}}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 448, "padding_multiple": 448, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 448, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 72, "inner_partitions": 10, "result_row_partitions": 10, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 10, "padding_groups": 1, "block_size": 448, "padding_multiple": 448, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 448, "column_block": 16}}}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 448, "padding_multiple": 448, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 736, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 72, "inner_partitions": 6, "result_row_partitions": 6, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 736, "padding_multiple": 736, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 736, "column_block": 16}}}, "tiling": {"tile_count": 432, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 736, "padding_multiple": 736, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 736, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 72, "inner_partitions": 6, "result_row_partitions": 6, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 736, "padding_multiple": 736, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 736, "column_block": 16}}}, "tiling": {"tile_count": 432, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 736, "padding_multiple": 736, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 1440, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 6, "column_partitions": 72, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 1440, "padding_multiple": 1440, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 1440, "column_block": 16}}}, "tiling": {"tile_count": 216, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 1440, "padding_multiple": 1440, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 1440, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 6, "column_partitions": 72, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 1440, "padding_multiple": 1440, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 1440, "column_block": 16}}}, "tiling": {"tile_count": 216, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 1440, "padding_multiple": 1440, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 544, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 36, "inner_partitions": 8, "result_row_partitions": 4, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 544, "padding_multiple": 544, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 544, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 544, "padding_multiple": 544, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 544, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 36, "inner_partitions": 8, "result_row_partitions": 4, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 544, "padding_multiple": 544, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 544, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 544, "padding_multiple": 544, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 544, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 36, "inner_partitions": 8, "result_row_partitions": 8, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 544, "padding_multiple": 544, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 544, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 544, "padding_multiple": 544, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 40, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 36}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 544, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 36, "inner_partitions": 8, "result_row_partitions": 8, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 544, "padding_multiple": 544, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 544, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 544, "padding_multiple": 544, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 40, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 36}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 480, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 72, "inner_partitions": 9, "result_row_partitions": 9, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 480, "padding_multiple": 480, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 480, "column_block": 16}}}, "tiling": {"tile_count": 648, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 480, "padding_multiple": 480, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 480, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 72, "inner_partitions": 9, "result_row_partitions": 9, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 480, "padding_multiple": 480, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 480, "column_block": 16}}}, "tiling": {"tile_count": 648, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 480, "padding_multiple": 480, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 2176, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 72, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 2176, "padding_multiple": 2176, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 2176, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2176, "padding_multiple": 2176, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 2176, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 72, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 2176, "padding_multiple": 2176, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 2176, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2176, "padding_multiple": 2176, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 448, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 4, "column_partitions": 36, "inner_partitions": 10, "result_row_partitions": 5, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 10, "padding_groups": 1, "block_size": 448, "padding_multiple": 448, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 448, "column_block": 16}}}, "tiling": {"tile_count": 360, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 448, "padding_multiple": 448, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 448, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 4, "column_partitions": 36, "inner_partitions": 10, "result_row_partitions": 5, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 10, "padding_groups": 1, "block_size": 448, "padding_multiple": 448, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 448, "column_block": 16}}}, "tiling": {"tile_count": 360, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 448, "padding_multiple": 448, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 2176, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 72, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 2176, "padding_multiple": 2176, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 2176, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2176, "padding_multiple": 2176, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 2176, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 72, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 2176, "padding_multiple": 2176, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 2176, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2176, "padding_multiple": 2176, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 1088, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 36, "inner_partitions": 4, "result_row_partitions": 2, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 1088, "padding_multiple": 1088, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 1088, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 1088, "padding_multiple": 1088, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 1088, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 36, "inner_partitions": 4, "result_row_partitions": 2, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 1088, "padding_multiple": 1088, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 1088, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 1088, "padding_multiple": 1088, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 640, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 42, "inner_partitions": 7, "result_row_partitions": 7, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 42, "axes": [{"axis": {"FromEnd": 1}, "partitions": 7, "padding_groups": 1, "block_size": 640, "padding_multiple": 640, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 640, "column_block": 16}}}, "tiling": {"tile_count": 294, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 42, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 7, "padding_groups": 1, "block_size": 640, "padding_multiple": 640, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 42, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 35, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 42}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 640, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 42, "inner_partitions": 7, "result_row_partitions": 7, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 42, "axes": [{"axis": {"FromEnd": 1}, "partitions": 7, "padding_groups": 1, "block_size": 640, "padding_multiple": 640, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 640, "column_block": 16}}}, "tiling": {"tile_count": 294, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 42, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 7, "padding_groups": 1, "block_size": 640, "padding_multiple": 640, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 42, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 35, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 42}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 352, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 37, "inner_partitions": 13, "result_row_partitions": 13, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1443, "replicas": 37, "axes": [{"axis": {"FromEnd": 1}, "partitions": 13, "padding_groups": 1, "block_size": 352, "padding_multiple": 352, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 352, "column_block": 16}}}, "tiling": {"tile_count": 481, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 37, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 13, "padding_groups": 1, "block_size": 352, "padding_multiple": 352, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1443, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 37, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 39, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 37}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 352, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 37, "inner_partitions": 13, "result_row_partitions": 13, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1443, "replicas": 37, "axes": [{"axis": {"FromEnd": 1}, "partitions": 13, "padding_groups": 1, "block_size": 352, "padding_multiple": 352, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 352, "column_block": 16}}}, "tiling": {"tile_count": 481, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 37, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 13, "padding_groups": 1, "block_size": 352, "padding_multiple": 352, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1443, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 37, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 39, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 37}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 736, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 24, "inner_partitions": 6, "result_row_partitions": 2, "result_column_partitions": 3, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 24, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 736, "padding_multiple": 736, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 736, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 736, "padding_multiple": 736, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 736, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 24, "inner_partitions": 6, "result_row_partitions": 2, "result_column_partitions": 3, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 24, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 736, "padding_multiple": 736, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 736, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 736, "padding_multiple": 736, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 4, "column_partitions": 24, "inner_partitions": 15, "result_row_partitions": 5, "result_column_partitions": 3, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 24, "axes": [{"axis": {"FromEnd": 1}, "partitions": 15, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 360, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 15, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 4, "column_partitions": 24, "inner_partitions": 15, "result_row_partitions": 5, "result_column_partitions": 3, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 24, "axes": [{"axis": {"FromEnd": 1}, "partitions": 15, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 360, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 15, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 24, "inner_partitions": 12, "result_row_partitions": 4, "result_column_partitions": 3, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 24, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 24, "inner_partitions": 12, "result_row_partitions": 4, "result_column_partitions": 3, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 24, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 1440, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 6, "column_partitions": 72, "inner_partitions": 3, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 1440, "padding_multiple": 1440, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 1440, "column_block": 16}}}, "tiling": {"tile_count": 216, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 1440, "padding_multiple": 1440, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 432, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 1440, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 6, "column_partitions": 72, "inner_partitions": 3, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 1440, "padding_multiple": 1440, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 1440, "column_block": 16}}}, "tiling": {"tile_count": 216, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 1440, "padding_multiple": 1440, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 432, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 224, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 20, "result_row_partitions": 20, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 20, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 224, "column_block": 16}}}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 224, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 20, "result_row_partitions": 20, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 20, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 224, "column_block": 16}}}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 24, "inner_partitions": 12, "result_row_partitions": 12, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 24, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 60, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 24}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 24, "inner_partitions": 12, "result_row_partitions": 12, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 24, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 60, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 24}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 1088, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 72, "inner_partitions": 4, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 1088, "padding_multiple": 1088, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 1088, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 1088, "padding_multiple": 1088, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 360, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 1088, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 72, "inner_partitions": 4, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 1088, "padding_multiple": 1088, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 1088, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 1088, "padding_multiple": 1088, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 360, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 4, "column_partitions": 24, "inner_partitions": 15, "result_row_partitions": 15, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 24, "axes": [{"axis": {"FromEnd": 1}, "partitions": 15, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 360, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 15, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 60, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 24}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 4, "column_partitions": 24, "inner_partitions": 15, "result_row_partitions": 15, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 24, "axes": [{"axis": {"FromEnd": 1}, "partitions": 15, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 360, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 15, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 60, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 24}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 416, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 26, "inner_partitions": 11, "result_row_partitions": 11, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1430, "replicas": 26, "axes": [{"axis": {"FromEnd": 1}, "partitions": 11, "padding_groups": 1, "block_size": 416, "padding_multiple": 416, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 416, "column_block": 16}}}, "tiling": {"tile_count": 286, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 26, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 11, "padding_groups": 1, "block_size": 416, "padding_multiple": 416, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1430, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 26, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 55, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 26}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 416, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 26, "inner_partitions": 11, "result_row_partitions": 11, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1430, "replicas": 26, "axes": [{"axis": {"FromEnd": 1}, "partitions": 11, "padding_groups": 1, "block_size": 416, "padding_multiple": 416, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 416, "column_block": 16}}}, "tiling": {"tile_count": 286, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 26, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 11, "padding_groups": 1, "block_size": 416, "padding_multiple": 416, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1430, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 26, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 55, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 26}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 224, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 36, "inner_partitions": 20, "result_row_partitions": 10, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 20, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 224, "column_block": 16}}}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 224, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 36, "inner_partitions": 20, "result_row_partitions": 10, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 20, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 224, "column_block": 16}}}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 4, "column_partitions": 26, "inner_partitions": 14, "result_row_partitions": 14, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1456, "replicas": 26, "axes": [{"axis": {"FromEnd": 1}, "partitions": 14, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 364, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 26, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 14, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1456, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 26, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 56, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 26}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 4, "column_partitions": 26, "inner_partitions": 14, "result_row_partitions": 14, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1456, "replicas": 26, "axes": [{"axis": {"FromEnd": 1}, "partitions": 14, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 364, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 26, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 14, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1456, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 26, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 56, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 26}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 17, "result_row_partitions": 17, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1224, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 17, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 1224, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 17, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1224, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 17, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 17, "result_row_partitions": 17, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1224, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 17, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 1224, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 17, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1224, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 17, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 480, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 9, "column_partitions": 18, "inner_partitions": 9, "result_row_partitions": 3, "result_column_partitions": 3, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 480, "padding_multiple": 480, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 480, "column_block": 16}}}, "tiling": {"tile_count": 162, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 480, "padding_multiple": 480, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 27, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 54}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 480, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 9, "column_partitions": 18, "inner_partitions": 9, "result_row_partitions": 3, "result_column_partitions": 3, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 480, "padding_multiple": 480, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 480, "column_block": 16}}}, "tiling": {"tile_count": 162, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 480, "padding_multiple": 480, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 27, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 54}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 80, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 7, "column_partitions": 15, "inner_partitions": 14, "result_row_partitions": 7, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 15, "axes": [{"axis": {"FromEnd": 1}, "partitions": 14, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 7, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 210, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 15, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 14, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 30, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 49, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 30}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 80, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 7, "column_partitions": 15, "inner_partitions": 14, "result_row_partitions": 7, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 15, "axes": [{"axis": {"FromEnd": 1}, "partitions": 14, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 7, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 210, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 15, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 14, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 30, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 49, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 30}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 864, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 4, "column_partitions": 72, "inner_partitions": 5, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 864, "padding_multiple": 864, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 864, "column_block": 16}}}, "tiling": {"tile_count": 360, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 864, "padding_multiple": 864, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 864, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 4, "column_partitions": 72, "inner_partitions": 5, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 864, "padding_multiple": 864, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 864, "column_block": 16}}}, "tiling": {"tile_count": 360, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 864, "padding_multiple": 864, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 2176, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 36, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 2176, "padding_multiple": 2176, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 2176, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2176, "padding_multiple": 2176, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 2176, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 36, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 2176, "padding_multiple": 2176, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 2176, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2176, "padding_multiple": 2176, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 1440, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 24, "inner_partitions": 3, "result_row_partitions": 1, "result_column_partitions": 3, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 24, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 1440, "padding_multiple": 1440, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 1440, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 1440, "padding_multiple": 1440, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 1440, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 24, "inner_partitions": 3, "result_row_partitions": 1, "result_column_partitions": 3, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 24, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 1440, "padding_multiple": 1440, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 1440, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 1440, "padding_multiple": 1440, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 1088, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 18, "inner_partitions": 4, "result_row_partitions": 1, "result_column_partitions": 4, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 1088, "padding_multiple": 1088, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 1088, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 1088, "padding_multiple": 1088, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 1088, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 18, "inner_partitions": 4, "result_row_partitions": 1, "result_column_partitions": 4, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 1088, "padding_multiple": 1088, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 1088, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 1088, "padding_multiple": 1088, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 18, "inner_partitions": 27, "result_row_partitions": 27, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 27, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 486, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 27, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 81, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 18, "inner_partitions": 27, "result_row_partitions": 27, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 27, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 486, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 27, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 81, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 144, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 8, "column_partitions": 8, "inner_partitions": 23, "result_row_partitions": 23, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 8, "axes": [{"axis": {"FromEnd": 1}, "partitions": 23, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 184, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 23, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 184, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 8}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 144, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 8, "column_partitions": 8, "inner_partitions": 23, "result_row_partitions": 23, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 8, "axes": [{"axis": {"FromEnd": 1}, "partitions": 23, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 184, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 23, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 184, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 8}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 736, "output_column_block": 96, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 12, "inner_partitions": 6, "result_row_partitions": 1, "result_column_partitions": 6, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 12, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 736, "padding_multiple": 736, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 736, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 736, "padding_multiple": 736, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 736, "output_column_block": 96, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 12, "inner_partitions": 6, "result_row_partitions": 1, "result_column_partitions": 6, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 12, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 736, "padding_multiple": 736, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 736, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 736, "padding_multiple": 736, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 544, "output_column_block": 128, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 9, "inner_partitions": 8, "result_row_partitions": 1, "result_column_partitions": 8, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 544, "padding_multiple": 544, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 544, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 544, "padding_multiple": 544, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 544, "output_column_block": 128, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 9, "inner_partitions": 8, "result_row_partitions": 1, "result_column_partitions": 8, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 544, "padding_multiple": 544, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 544, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 544, "padding_multiple": 544, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 480, "output_column_block": 144, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 8, "inner_partitions": 9, "result_row_partitions": 1, "result_column_partitions": 9, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 8, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 480, "padding_multiple": 480, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 480, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 480, "padding_multiple": 480, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 480, "output_column_block": 144, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 8, "inner_partitions": 9, "result_row_partitions": 1, "result_column_partitions": 9, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 8, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 480, "padding_multiple": 480, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 480, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 480, "padding_multiple": 480, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 128, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 6, "column_partitions": 9, "inner_partitions": 27, "result_row_partitions": 9, "result_column_partitions": 3, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 27, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 243, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 27, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 27, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 54, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 27}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 128, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 6, "column_partitions": 9, "inner_partitions": 27, "result_row_partitions": 9, "result_column_partitions": 3, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 27, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 243, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 27, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 27, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 54, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 27}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 864, "output_column_block": 96, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 24, "column_partitions": 12, "inner_partitions": 5, "result_row_partitions": 1, "result_column_partitions": 5, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 12, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 864, "padding_multiple": 864, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 24, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 864, "column_block": 16}}}, "tiling": {"tile_count": 60, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 864, "padding_multiple": 864, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 60, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 24, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 60}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 864, "output_column_block": 96, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 24, "column_partitions": 12, "inner_partitions": 5, "result_row_partitions": 1, "result_column_partitions": 5, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 12, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 864, "padding_multiple": 864, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 24, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 864, "column_block": 16}}}, "tiling": {"tile_count": 60, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 864, "padding_multiple": 864, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 60, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 24, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 60}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 192, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 6, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 12, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 192, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 6, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 12, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 640, "output_column_block": 128, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 23, "column_partitions": 9, "inner_partitions": 7, "result_row_partitions": 1, "result_column_partitions": 7, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1449, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 7, "padding_groups": 1, "block_size": 640, "padding_multiple": 640, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 23, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 640, "column_block": 16}}}, "tiling": {"tile_count": 63, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 7, "padding_groups": 1, "block_size": 640, "padding_multiple": 640, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1449, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 63, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 23, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 63}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 640, "output_column_block": 128, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 23, "column_partitions": 9, "inner_partitions": 7, "result_row_partitions": 1, "result_column_partitions": 7, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1449, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 7, "padding_groups": 1, "block_size": 640, "padding_multiple": 640, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 23, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 640, "column_block": 16}}}, "tiling": {"tile_count": 63, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 7, "padding_groups": 1, "block_size": 640, "padding_multiple": 640, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1449, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 63, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 23, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 63}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 43, "inner_partitions": 34, "result_row_partitions": 34, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1462, "replicas": 43, "axes": [{"axis": {"FromEnd": 1}, "partitions": 34, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 1462, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 43, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 34, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1462, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 43, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 34, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 43}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 43, "inner_partitions": 34, "result_row_partitions": 34, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1462, "replicas": 43, "axes": [{"axis": {"FromEnd": 1}, "partitions": 34, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 1462, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 43, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 34, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1462, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 43, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 34, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 43}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 736, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 72, "inner_partitions": 6, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 736, "padding_multiple": 736, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 736, "column_block": 16}}}, "tiling": {"tile_count": 432, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 736, "padding_multiple": 736, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 216, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 736, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 72, "inner_partitions": 6, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 736, "padding_multiple": 736, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 736, "column_block": 16}}}, "tiling": {"tile_count": 432, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 736, "padding_multiple": 736, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 216, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 144, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 4, "column_partitions": 8, "inner_partitions": 45, "result_row_partitions": 5, "result_column_partitions": 9, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 8, "axes": [{"axis": {"FromEnd": 1}, "partitions": 45, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 360, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 45, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 144, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 4, "column_partitions": 8, "inner_partitions": 45, "result_row_partitions": 5, "result_column_partitions": 9, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 8, "axes": [{"axis": {"FromEnd": 1}, "partitions": 45, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 360, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 45, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 448, "output_column_block": 192, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 24, "column_partitions": 6, "inner_partitions": 10, "result_row_partitions": 1, "result_column_partitions": 10, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 10, "padding_groups": 1, "block_size": 448, "padding_multiple": 448, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 24, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 448, "column_block": 16}}}, "tiling": {"tile_count": 60, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 448, "padding_multiple": 448, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 60, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 24, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 60}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 448, "output_column_block": 192, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 24, "column_partitions": 6, "inner_partitions": 10, "result_row_partitions": 1, "result_column_partitions": 10, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 10, "padding_groups": 1, "block_size": 448, "padding_multiple": 448, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 24, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 448, "column_block": 16}}}, "tiling": {"tile_count": 60, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 448, "padding_multiple": 448, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 60, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 24, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 60}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 416, "output_column_block": 192, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 22, "column_partitions": 6, "inner_partitions": 11, "result_row_partitions": 1, "result_column_partitions": 11, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1452, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 11, "padding_groups": 1, "block_size": 416, "padding_multiple": 416, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 22, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 416, "column_block": 16}}}, "tiling": {"tile_count": 66, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 11, "padding_groups": 1, "block_size": 416, "padding_multiple": 416, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1452, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 66, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 22, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 66}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 416, "output_column_block": 192, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 22, "column_partitions": 6, "inner_partitions": 11, "result_row_partitions": 1, "result_column_partitions": 11, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1452, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 11, "padding_groups": 1, "block_size": 416, "padding_multiple": 416, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 22, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 416, "column_block": 16}}}, "tiling": {"tile_count": 66, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 11, "padding_groups": 1, "block_size": 416, "padding_multiple": 416, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1452, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 66, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 22, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 66}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 192, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 7, "column_partitions": 6, "inner_partitions": 34, "result_row_partitions": 17, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1428, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 34, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 7, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 204, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 34, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1428, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 119, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 12}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 192, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 7, "column_partitions": 6, "inner_partitions": 34, "result_row_partitions": 17, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1428, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 34, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 7, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 204, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 34, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1428, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 119, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 12}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 240, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 21, "column_partitions": 5, "inner_partitions": 14, "result_row_partitions": 1, "result_column_partitions": 14, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 5, "axes": [{"axis": {"FromEnd": 1}, "partitions": 14, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 70, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 14, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 70, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 70}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 240, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 21, "column_partitions": 5, "inner_partitions": 14, "result_row_partitions": 1, "result_column_partitions": 14, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 5, "axes": [{"axis": {"FromEnd": 1}, "partitions": 14, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 70, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 14, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 70, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 70}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 352, "output_column_block": 288, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 28, "column_partitions": 4, "inner_partitions": 13, "result_row_partitions": 1, "result_column_partitions": 13, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1456, "replicas": 4, "axes": [{"axis": {"FromEnd": 1}, "partitions": 13, "padding_groups": 1, "block_size": 352, "padding_multiple": 352, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 28, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 352, "column_block": 16}}}, "tiling": {"tile_count": 52, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 13, "padding_groups": 1, "block_size": 352, "padding_multiple": 352, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1456, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 52, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 28, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 52}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 352, "output_column_block": 288, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 28, "column_partitions": 4, "inner_partitions": 13, "result_row_partitions": 1, "result_column_partitions": 13, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1456, "replicas": 4, "axes": [{"axis": {"FromEnd": 1}, "partitions": 13, "padding_groups": 1, "block_size": 352, "padding_multiple": 352, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 28, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 352, "column_block": 16}}}, "tiling": {"tile_count": 52, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 13, "padding_groups": 1, "block_size": 352, "padding_multiple": 352, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1456, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 52, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 28, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 52}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 288, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 21, "column_partitions": 4, "inner_partitions": 17, "result_row_partitions": 1, "result_column_partitions": 17, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1428, "replicas": 4, "axes": [{"axis": {"FromEnd": 1}, "partitions": 17, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 68, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 17, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1428, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 68, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 68}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 288, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 21, "column_partitions": 4, "inner_partitions": 17, "result_row_partitions": 1, "result_column_partitions": 17, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1428, "replicas": 4, "axes": [{"axis": {"FromEnd": 1}, "partitions": 17, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 68, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 17, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1428, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 68, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 68}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 384, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 32, "column_partitions": 3, "inner_partitions": 15, "result_row_partitions": 1, "result_column_partitions": 15, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 3, "axes": [{"axis": {"FromEnd": 1}, "partitions": 15, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 45, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 15, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 45, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 45}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 384, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 32, "column_partitions": 3, "inner_partitions": 15, "result_row_partitions": 1, "result_column_partitions": 15, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 3, "axes": [{"axis": {"FromEnd": 1}, "partitions": 15, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 45, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 15, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 45, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 45}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 640, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 8, "column_partitions": 26, "inner_partitions": 7, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1456, "replicas": 26, "axes": [{"axis": {"FromEnd": 1}, "partitions": 7, "padding_groups": 1, "block_size": 640, "padding_multiple": 640, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 640, "column_block": 16}}}, "tiling": {"tile_count": 182, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 26, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 7, "padding_groups": 1, "block_size": 640, "padding_multiple": 640, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 208, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 26, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 26}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 640, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 8, "column_partitions": 26, "inner_partitions": 7, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1456, "replicas": 26, "axes": [{"axis": {"FromEnd": 1}, "partitions": 7, "padding_groups": 1, "block_size": 640, "padding_multiple": 640, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 640, "column_block": 16}}}, "tiling": {"tile_count": 182, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 26, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 7, "padding_groups": 1, "block_size": 640, "padding_multiple": 640, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 208, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 26, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 26}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 544, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 36, "inner_partitions": 8, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 544, "padding_multiple": 544, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 544, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 544, "padding_multiple": 544, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 180, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 36}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 544, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 36, "inner_partitions": 8, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 544, "padding_multiple": 544, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 544, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 544, "padding_multiple": 544, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 180, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 36}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 144, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 4, "column_partitions": 8, "inner_partitions": 45, "result_row_partitions": 45, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 8, "axes": [{"axis": {"FromEnd": 1}, "partitions": 45, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 360, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 45, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 180, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 8}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 144, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 4, "column_partitions": 8, "inner_partitions": 45, "result_row_partitions": 45, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 8, "axes": [{"axis": {"FromEnd": 1}, "partitions": 45, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 360, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 45, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 180, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 8}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 576, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 32, "column_partitions": 2, "inner_partitions": 23, "result_row_partitions": 1, "result_column_partitions": 23, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 2, "axes": [{"axis": {"FromEnd": 1}, "partitions": 23, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 46, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 23, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 46, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 46}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 576, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 32, "column_partitions": 2, "inner_partitions": 23, "result_row_partitions": 1, "result_column_partitions": 23, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 2, "axes": [{"axis": {"FromEnd": 1}, "partitions": 23, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 46, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 23, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 46, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 46}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 576, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 27, "column_partitions": 2, "inner_partitions": 27, "result_row_partitions": 1, "result_column_partitions": 27, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 2, "axes": [{"axis": {"FromEnd": 1}, "partitions": 27, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 27, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 54, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 27, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 27, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 54}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 576, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 27, "column_partitions": 2, "inner_partitions": 27, "result_row_partitions": 1, "result_column_partitions": 27, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 2, "axes": [{"axis": {"FromEnd": 1}, "partitions": 27, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 27, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 54, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 27, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 27, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 54}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 480, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 9, "column_partitions": 18, "inner_partitions": 9, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 480, "padding_multiple": 480, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 480, "column_block": 16}}}, "tiling": {"tile_count": 162, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 480, "padding_multiple": 480, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 162, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 480, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 9, "column_partitions": 18, "inner_partitions": 9, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 480, "padding_multiple": 480, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 480, "column_block": 16}}}, "tiling": {"tile_count": 162, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 480, "padding_multiple": 480, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 162, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 576, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 21, "column_partitions": 2, "inner_partitions": 34, "result_row_partitions": 1, "result_column_partitions": 34, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1428, "replicas": 2, "axes": [{"axis": {"FromEnd": 1}, "partitions": 34, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 68, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 34, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1428, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 68, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 68}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 576, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 21, "column_partitions": 2, "inner_partitions": 34, "result_row_partitions": 1, "result_column_partitions": 34, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1428, "replicas": 2, "axes": [{"axis": {"FromEnd": 1}, "partitions": 34, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 68, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 34, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1428, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 68, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 68}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 448, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 72, "inner_partitions": 10, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 10, "padding_groups": 1, "block_size": 448, "padding_multiple": 448, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 448, "column_block": 16}}}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 448, "padding_multiple": 448, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 448, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 72, "inner_partitions": 10, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 10, "padding_groups": 1, "block_size": 448, "padding_multiple": 448, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 448, "column_block": 16}}}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 448, "padding_multiple": 448, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 224, "output_column_block": 1152, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 73, "column_partitions": 1, "inner_partitions": 20, "result_row_partitions": 1, "result_column_partitions": 20, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 20, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 73, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 224, "column_block": 16}}}, "tiling": {"tile_count": 20, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 20, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 73, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 20}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 224, "output_column_block": 1152, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 73, "column_partitions": 1, "inner_partitions": 20, "result_row_partitions": 1, "result_column_partitions": 20, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 20, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 73, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 224, "column_block": 16}}}, "tiling": {"tile_count": 20, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 20, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 73, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 20}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 416, "output_column_block": 96, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 11, "column_partitions": 12, "inner_partitions": 11, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1452, "replicas": 12, "axes": [{"axis": {"FromEnd": 1}, "partitions": 11, "padding_groups": 1, "block_size": 416, "padding_multiple": 416, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 11, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 416, "column_block": 16}}}, "tiling": {"tile_count": 132, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 11, "padding_groups": 1, "block_size": 416, "padding_multiple": 416, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 132, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 11, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 12}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 416, "output_column_block": 96, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 11, "column_partitions": 12, "inner_partitions": 11, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1452, "replicas": 12, "axes": [{"axis": {"FromEnd": 1}, "partitions": 11, "padding_groups": 1, "block_size": 416, "padding_multiple": 416, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 11, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 416, "column_block": 16}}}, "tiling": {"tile_count": 132, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 11, "padding_groups": 1, "block_size": 416, "padding_multiple": 416, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 132, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 11, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 12}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 576, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 21, "column_partitions": 2, "inner_partitions": 34, "result_row_partitions": 2, "result_column_partitions": 17, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1428, "replicas": 2, "axes": [{"axis": {"FromEnd": 1}, "partitions": 34, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 68, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 34, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1428, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 34, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 42, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 34}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 576, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 21, "column_partitions": 2, "inner_partitions": 34, "result_row_partitions": 2, "result_column_partitions": 17, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1428, "replicas": 2, "axes": [{"axis": {"FromEnd": 1}, "partitions": 34, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 68, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 34, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1428, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 34, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 42, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 34}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 24, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 24, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 120, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 24}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 24, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 24, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 120, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 24}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 352, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 37, "inner_partitions": 13, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1443, "replicas": 37, "axes": [{"axis": {"FromEnd": 1}, "partitions": 13, "padding_groups": 1, "block_size": 352, "padding_multiple": 352, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 352, "column_block": 16}}}, "tiling": {"tile_count": 481, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 37, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 13, "padding_groups": 1, "block_size": 352, "padding_multiple": 352, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 111, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 37, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 37}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 352, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 37, "inner_partitions": 13, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1443, "replicas": 37, "axes": [{"axis": {"FromEnd": 1}, "partitions": 13, "padding_groups": 1, "block_size": 352, "padding_multiple": 352, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 352, "column_block": 16}}}, "tiling": {"tile_count": 481, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 37, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 13, "padding_groups": 1, "block_size": 352, "padding_multiple": 352, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 111, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 37, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 37}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}], "22": [{"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1458, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1472, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1472, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1472, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 729, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1024, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1024, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1024, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 512}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 512, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 512}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 256}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 256, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 256}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 128}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 128, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 128}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 64}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 64, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 64}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 32, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 16}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 16, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 16}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 8}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 8, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 8}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}], "23": [{"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1472, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1472, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1472, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1024, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1024, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1024, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 512}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 512}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 512}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 256}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 256}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 256}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 128}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 128}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 128}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 64}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 64}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 64}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 16}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 16}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 16}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 8}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 8}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 8}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}], "25": [{"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1458, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1458, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1472, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1472, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1472, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1472, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 729, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 729, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1024, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1024, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1024, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1024, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 512}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 512, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 512, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 512}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 256}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 256, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 256, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 256}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 128}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 128, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 128, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 128}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 64}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 64, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 64, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 64}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 32, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 32, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 16}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 16, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 16, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 16}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 8}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 8, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 8, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 8}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}], "26": [{"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 144, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 144, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 144}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 144, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 144, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 144}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 144, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 144, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 144}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 144, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 144, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 144}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 144, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 144, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 144}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 144, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 144, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 144}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 144, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 144, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 144}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 144, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 144, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 144}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 144, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 144, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 432, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 144}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 144, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 144, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 432, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 144}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 144, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 144, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 432, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 144}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 144, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 144, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 432, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 144}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 144, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 144, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 144}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 144, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 144, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 144}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 144, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 144, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 144}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 144, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 144, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 144}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 144, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 144, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 5}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 144, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 144, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 5}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 144, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 144, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 5}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 144, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 144, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 5}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 73, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 73, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 146, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 73}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 73, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 73, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 146, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 73}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 73, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 73, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 146, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 73}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 73, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 73, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 146, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 73}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 73, "inner_partitions": 4, "result_row_partitions": 4, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 73, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 292, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 73}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 73, "inner_partitions": 4, "result_row_partitions": 4, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 73, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 292, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 73}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 144, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 144, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 144}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 144, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 144, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 144}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 144, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 144, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 144}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 144, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 144, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 144}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 144, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 144, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 5}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 144, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 144, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 5}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 144, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 144, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 5}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 144, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 144, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 5}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 73, "inner_partitions": 4, "result_row_partitions": 4, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 73, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 292, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 73}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 73, "inner_partitions": 4, "result_row_partitions": 4, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 73, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 292, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 73}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 144, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 144, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 432, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 144}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 144, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 144, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 432, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 144}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 144, "inner_partitions": 5, "result_row_partitions": 5, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 144, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 144}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 144, "inner_partitions": 5, "result_row_partitions": 5, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 144, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 144}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 144, "inner_partitions": 5, "result_row_partitions": 5, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 144, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 144}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 144, "inner_partitions": 5, "result_row_partitions": 5, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 144, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 144}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 73, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 73, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 146, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 73}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 73, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 73, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 146, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 73}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 73, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 73, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 146, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 73}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 73, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 73, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 146, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 73}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 49, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 49, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 147, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 49, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 49, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 30, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 49}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 49, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 49, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 147, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 49, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 49, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 30, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 49}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 144, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 144, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 432, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 144}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 144, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 144, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 432, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 144}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 4, "column_partitions": 73, "inner_partitions": 5, "result_row_partitions": 5, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 73, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 365, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 73}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 4, "column_partitions": 73, "inner_partitions": 5, "result_row_partitions": 5, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 73, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 365, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 73}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 73, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 73, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 146, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 730, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 73}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 73, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 73, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 146, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 730, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 73}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 73, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 73, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 146, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 730, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 73}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 73, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 73, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 146, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 730, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 73}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 73, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 73, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 146, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 730, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 10}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 73, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 73, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 146, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 730, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 10}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 73, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 73, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 146, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 730, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 10}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 73, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 73, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 146, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 730, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 10}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 15, "column_partitions": 49, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 49, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 15, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 98, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 49, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 49, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 30, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 49}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 15, "column_partitions": 49, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 49, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 15, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 98, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 49, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 49, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 30, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 49}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 15, "column_partitions": 49, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 49, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 15, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 98, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 49, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 49, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 30, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 49}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 15, "column_partitions": 49, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 49, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 15, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 98, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 49, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 49, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 30, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 49}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 36, "inner_partitions": 4, "result_row_partitions": 1, "result_column_partitions": 4, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 144}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 36, "inner_partitions": 4, "result_row_partitions": 1, "result_column_partitions": 4, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 144}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 36, "inner_partitions": 4, "result_row_partitions": 1, "result_column_partitions": 4, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 144}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 36, "inner_partitions": 4, "result_row_partitions": 1, "result_column_partitions": 4, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 144}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 7, "column_partitions": 52, "inner_partitions": 4, "result_row_partitions": 4, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1456, "replicas": 52, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 7, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 208, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 52, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1456, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 52, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 28, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 52}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 7, "column_partitions": 52, "inner_partitions": 4, "result_row_partitions": 4, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1456, "replicas": 52, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 7, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 208, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 52, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1456, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 52, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 28, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 52}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 36, "inner_partitions": 4, "result_row_partitions": 2, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 36, "inner_partitions": 4, "result_row_partitions": 2, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 36, "inner_partitions": 4, "result_row_partitions": 2, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 36, "inner_partitions": 4, "result_row_partitions": 2, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 49, "inner_partitions": 6, "result_row_partitions": 6, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 49, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 294, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 49, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 49, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 30, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 49}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 49, "inner_partitions": 6, "result_row_partitions": 6, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 49, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 294, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 49, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 49, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 30, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 49}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 15, "column_partitions": 49, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 49, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 15, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 98, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 49, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 98, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 15, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 98}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 576, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 15, "column_partitions": 49, "inner_partitions": 2, "result_row_partitions": 1, "result_column_partitions": 2, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 49, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 15, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 576, "column_block": 16}}}, "tiling": {"tile_count": 98, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 49, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 576, "padding_multiple": 576, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 98, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 15, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 98}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 96, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 24, "inner_partitions": 6, "result_row_partitions": 1, "result_column_partitions": 6, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 24, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 144}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 96, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 24, "inner_partitions": 6, "result_row_partitions": 1, "result_column_partitions": 6, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 24, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 144}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 13, "column_partitions": 37, "inner_partitions": 3, "result_row_partitions": 1, "result_column_partitions": 3, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1443, "replicas": 37, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 13, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 111, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 37, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1443, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 111, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 13, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 111}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 13, "column_partitions": 37, "inner_partitions": 3, "result_row_partitions": 1, "result_column_partitions": 3, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1443, "replicas": 37, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 13, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 111, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 37, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1443, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 111, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 13, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 111}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 96, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 24, "inner_partitions": 6, "result_row_partitions": 2, "result_column_partitions": 3, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 24, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 96, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 24, "inner_partitions": 6, "result_row_partitions": 2, "result_column_partitions": 3, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 24, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 144, "inner_partitions": 9, "result_row_partitions": 9, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 144, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 144}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 144, "inner_partitions": 9, "result_row_partitions": 9, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 144, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 144}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 36, "inner_partitions": 8, "result_row_partitions": 2, "result_column_partitions": 4, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 144}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 36, "inner_partitions": 8, "result_row_partitions": 2, "result_column_partitions": 4, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 144}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 36, "inner_partitions": 8, "result_row_partitions": 8, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 40, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 36}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 36, "inner_partitions": 8, "result_row_partitions": 8, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 40, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 36}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 144, "inner_partitions": 3, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 144, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 432, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 432, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 144}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 3, "column_partitions": 144, "inner_partitions": 3, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 144, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 432, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 432, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 144}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 128, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 18, "inner_partitions": 8, "result_row_partitions": 1, "result_column_partitions": 8, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 144}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 128, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 18, "inner_partitions": 8, "result_row_partitions": 1, "result_column_partitions": 8, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 144}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 128, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 16, "column_partitions": 18, "inner_partitions": 5, "result_row_partitions": 1, "result_column_partitions": 5, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 90, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 90, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 90}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 128, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 16, "column_partitions": 18, "inner_partitions": 5, "result_row_partitions": 1, "result_column_partitions": 5, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 90, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 90, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 90}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 128, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 9, "column_partitions": 18, "inner_partitions": 9, "result_row_partitions": 3, "result_column_partitions": 3, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 162, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 27, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 54}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 128, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 9, "column_partitions": 18, "inner_partitions": 9, "result_row_partitions": 3, "result_column_partitions": 3, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 162, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 27, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 54}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 96, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 24, "inner_partitions": 12, "result_row_partitions": 2, "result_column_partitions": 6, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 24, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 144}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 96, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 24, "inner_partitions": 12, "result_row_partitions": 2, "result_column_partitions": 6, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 24, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 144}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 96, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 24, "inner_partitions": 12, "result_row_partitions": 12, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 24, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 60, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 24}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 96, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 24, "inner_partitions": 12, "result_row_partitions": 12, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 24, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 60, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 24}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 192, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 12, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 12, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 12, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 144}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 192, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 10, "column_partitions": 12, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 12, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 12, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 144, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 144}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 256, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 18, "column_partitions": 9, "inner_partitions": 9, "result_row_partitions": 1, "result_column_partitions": 9, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 81, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 81, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 81}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 256, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 18, "column_partitions": 9, "inner_partitions": 9, "result_row_partitions": 1, "result_column_partitions": 9, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 81, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 81, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 81}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 73, "inner_partitions": 4, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 73, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 292, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 365, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 73}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 73, "inner_partitions": 4, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1460, "replicas": 73, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 292, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 365, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 73, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 73}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 192, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 12, "inner_partitions": 6, "result_row_partitions": 1, "result_column_partitions": 6, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 12, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 192, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 20, "column_partitions": 12, "inner_partitions": 6, "result_row_partitions": 1, "result_column_partitions": 6, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 12, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 144, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 16, "inner_partitions": 18, "result_row_partitions": 2, "result_column_partitions": 9, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 16, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 16, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 144}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 144, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 16, "inner_partitions": 18, "result_row_partitions": 2, "result_column_partitions": 9, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 16, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 16, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 10, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 144}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 81, "inner_partitions": 18, "result_row_partitions": 18, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 81, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 81, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 81, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 81}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 81, "inner_partitions": 18, "result_row_partitions": 18, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 81, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 81, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 81, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 81}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 144, "inner_partitions": 5, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 144, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 144}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 144, "inner_partitions": 5, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 144, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 720, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 144}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 768, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 27, "column_partitions": 3, "inner_partitions": 18, "result_row_partitions": 1, "result_column_partitions": 18, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 3, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 27, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 54, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 27, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 54}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 768, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 27, "column_partitions": 3, "inner_partitions": 18, "result_row_partitions": 1, "result_column_partitions": 18, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 3, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 27, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 54, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 27, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 54}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 49, "inner_partitions": 6, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 49, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 294, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 49, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 245, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 49, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 49}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 49, "inner_partitions": 6, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 49, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 294, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 49, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 245, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 49, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 49}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 36, "inner_partitions": 8, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 180, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 36}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 36, "inner_partitions": 8, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 36, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 180, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 36}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 128, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 9, "column_partitions": 18, "inner_partitions": 9, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 162, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 162, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 128, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 9, "column_partitions": 18, "inner_partitions": 9, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 162, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 162, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 96, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 24, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 24, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 120, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 24}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 96, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 24, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 24, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 120, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 24, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 24}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 144, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 16, "inner_partitions": 18, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 16, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 16, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 80, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 16, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 16}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 144, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 5, "column_partitions": 16, "inner_partitions": 18, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 16, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 288, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 16, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 80, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 16, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 16}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}], "27": [{"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1458, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1472, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1472, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1472, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 729, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 729}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1024, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1024, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1024, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 512}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 512, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 512}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 256}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 256, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 256}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 128}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 128, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 128}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 64}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 64, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 64}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 32, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 16}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 16, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 16}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}], "28": [{"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 864, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 864, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 864, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 864, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 864, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 864, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 864, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 864, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 18, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 18, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 18, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 18, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 245, "column_partitions": 1, "inner_partitions": 6, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 1470, "replicas": 245, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 245, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 245, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 245, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 245, "column_partitions": 1, "inner_partitions": 6, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 1470, "replicas": 245, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 245, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 245, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 245, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 245, "column_partitions": 1, "inner_partitions": 6, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 1470, "replicas": 245, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 245, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 245, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 245, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 245, "column_partitions": 1, "inner_partitions": 6, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 1470, "replicas": 245, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 245, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 245, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 245, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 294, "column_partitions": 1, "inner_partitions": 5, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 1470, "replicas": 294, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 294, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 294, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 294, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 294, "column_partitions": 1, "inner_partitions": 5, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 1470, "replicas": 294, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 294, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 294, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 294, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 294, "column_partitions": 1, "inner_partitions": 5, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 1470, "replicas": 294, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 294, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 294, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 294, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 294, "column_partitions": 1, "inner_partitions": 5, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 1470, "replicas": 294, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 294, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 294, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 294, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 163, "column_partitions": 1, "inner_partitions": 9, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 1467, "replicas": 163, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1467, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 163, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 163, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 163, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 163, "column_partitions": 1, "inner_partitions": 9, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 1467, "replicas": 163, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1467, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 163, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 163, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 163, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 163, "column_partitions": 1, "inner_partitions": 9, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 1467, "replicas": 163, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1467, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 163, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 163, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 163, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 163, "column_partitions": 1, "inner_partitions": 9, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 1467, "replicas": 163, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1467, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 163, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 163, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 163, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 368, "column_partitions": 1, "inner_partitions": 4, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 1472, "replicas": 368, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 368, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 368, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 368, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 368, "column_partitions": 1, "inner_partitions": 4, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 1472, "replicas": 368, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 368, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 368, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 368, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 368, "column_partitions": 1, "inner_partitions": 4, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 1472, "replicas": 368, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 368, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 368, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 368, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 368, "column_partitions": 1, "inner_partitions": 4, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 1472, "replicas": 368, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 368, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 368, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 368, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 245, "column_partitions": 1, "inner_partitions": 6, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 1470, "replicas": 245, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 245, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 245, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 245, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 245, "column_partitions": 1, "inner_partitions": 6, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 1470, "replicas": 245, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 245, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 245, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 245, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 245, "column_partitions": 1, "inner_partitions": 6, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 1470, "replicas": 245, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 245, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 245, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 245, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 245, "column_partitions": 1, "inner_partitions": 6, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 1470, "replicas": 245, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 245, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 245, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 245, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 184, "column_partitions": 1, "inner_partitions": 8, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 1472, "replicas": 184, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 184, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 184, "column_partitions": 1, "inner_partitions": 8, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 1472, "replicas": 184, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 184, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 184, "column_partitions": 1, "inner_partitions": 8, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 1472, "replicas": 184, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 184, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 184, "column_partitions": 1, "inner_partitions": 8, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 1472, "replicas": 184, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 184, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 163, "column_partitions": 1, "inner_partitions": 9, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 1467, "replicas": 163, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1467, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 163, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 163, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 163, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 163, "column_partitions": 1, "inner_partitions": 9, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 1467, "replicas": 163, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1467, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 163, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 163, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 163, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 163, "column_partitions": 1, "inner_partitions": 9, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 1467, "replicas": 163, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1467, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 163, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 163, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 163, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 163, "column_partitions": 1, "inner_partitions": 9, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 1467, "replicas": 163, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1467, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 163, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 163, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 163, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 294, "column_partitions": 1, "inner_partitions": 5, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 1470, "replicas": 294, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 294, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 294, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 294, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 294, "column_partitions": 1, "inner_partitions": 5, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 1470, "replicas": 294, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 294, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 294, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 294, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 294, "column_partitions": 1, "inner_partitions": 5, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 1470, "replicas": 294, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 294, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 294, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 294, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 294, "column_partitions": 1, "inner_partitions": 5, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 1470, "replicas": 294, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 294, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 294, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 294, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 40, "inner_partitions": 36, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 40, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 40, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 40}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 40, "inner_partitions": 36, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 40, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 40, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 40}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 184, "column_partitions": 1, "inner_partitions": 8, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 1472, "replicas": 184, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 184, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 184, "column_partitions": 1, "inner_partitions": 8, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 1472, "replicas": 184, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 184, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 184, "column_partitions": 1, "inner_partitions": 8, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 1472, "replicas": 184, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 184, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 184, "column_partitions": 1, "inner_partitions": 8, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 1472, "replicas": 184, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 184, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 40, "inner_partitions": 36, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 40, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 40, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 40}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 40, "inner_partitions": 36, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 40, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 40, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 40}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 368, "column_partitions": 1, "inner_partitions": 4, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 1472, "replicas": 368, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 368, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 368, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 368, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 368, "column_partitions": 1, "inner_partitions": 4, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 1472, "replicas": 368, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 368, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 368, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 368, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 368, "column_partitions": 1, "inner_partitions": 4, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 1472, "replicas": 368, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 368, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 368, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 368, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 368, "column_partitions": 1, "inner_partitions": 4, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 1472, "replicas": 368, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 368, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 368, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 368, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 490, "column_partitions": 1, "inner_partitions": 3, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 1470, "replicas": 490, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 490, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 490, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 490, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 490, "column_partitions": 1, "inner_partitions": 3, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 1470, "replicas": 490, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 490, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 490, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 490, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 490, "column_partitions": 1, "inner_partitions": 3, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 1470, "replicas": 490, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 490, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 490, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 490, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 490, "column_partitions": 1, "inner_partitions": 3, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 1470, "replicas": 490, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 490, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 490, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 490, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 122, "column_partitions": 1, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 1464, "replicas": 122, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1464, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 122, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 122, "column_partitions": 1, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 1464, "replicas": 122, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1464, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 122, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 122, "column_partitions": 1, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 1464, "replicas": 122, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1464, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 122, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 122, "column_partitions": 1, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 1464, "replicas": 122, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1464, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 122, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 122, "column_partitions": 1, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 1464, "replicas": 122, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1464, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 122, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 122, "column_partitions": 1, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 1464, "replicas": 122, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1464, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 122, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 122, "column_partitions": 1, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 1464, "replicas": 122, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1464, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 122, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 122, "column_partitions": 1, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 1464, "replicas": 122, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1464, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 122, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 368, "column_partitions": 1, "inner_partitions": 4, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 1472, "replicas": 368, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 368, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 368, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 368, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 368, "column_partitions": 1, "inner_partitions": 4, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 1472, "replicas": 368, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 368, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 368, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 368, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 368, "column_partitions": 1, "inner_partitions": 4, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 1472, "replicas": 368, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 368, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 368, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 368, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 368, "column_partitions": 1, "inner_partitions": 4, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 1472, "replicas": 368, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 368, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 368, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 368, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 490, "column_partitions": 1, "inner_partitions": 3, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 1470, "replicas": 490, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 490, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 490, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 490, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 490, "column_partitions": 1, "inner_partitions": 3, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 1470, "replicas": 490, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 490, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 490, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 490, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 490, "column_partitions": 1, "inner_partitions": 3, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 1470, "replicas": 490, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 490, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 490, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 490, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 490, "column_partitions": 1, "inner_partitions": 3, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 1470, "replicas": 490, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 490, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 490, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 490, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 490, "column_partitions": 1, "inner_partitions": 3, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 1470, "replicas": 490, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 490, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 490, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 490, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 490, "column_partitions": 1, "inner_partitions": 3, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 1470, "replicas": 490, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 490, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 490, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 490, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 490, "column_partitions": 1, "inner_partitions": 3, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 1470, "replicas": 490, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 490, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 490, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 490, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 490, "column_partitions": 1, "inner_partitions": 3, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 1470, "replicas": 490, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 490, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 490, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 490, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 294, "column_partitions": 1, "inner_partitions": 5, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 1470, "replicas": 294, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 294, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 294, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 294, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 294, "column_partitions": 1, "inner_partitions": 5, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 1470, "replicas": 294, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 294, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 294, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 294, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 294, "column_partitions": 1, "inner_partitions": 5, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 1470, "replicas": 294, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 294, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 294, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 294, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 294, "column_partitions": 1, "inner_partitions": 5, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 1470, "replicas": 294, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 294, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 294, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 294, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 368, "column_partitions": 1, "inner_partitions": 4, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 1472, "replicas": 368, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 368, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 368, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 368, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 368, "column_partitions": 1, "inner_partitions": 4, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 1472, "replicas": 368, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 368, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 368, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 368, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 368, "column_partitions": 1, "inner_partitions": 4, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 1472, "replicas": 368, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 368, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 368, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 368, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 368, "column_partitions": 1, "inner_partitions": 4, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 1472, "replicas": 368, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 368, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 368, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 368, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 294, "column_partitions": 1, "inner_partitions": 5, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 1470, "replicas": 294, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 294, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 294, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 294, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 294, "column_partitions": 1, "inner_partitions": 5, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 1470, "replicas": 294, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 294, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 294, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 294, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 294, "column_partitions": 1, "inner_partitions": 5, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 1470, "replicas": 294, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 294, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 294, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 294, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 294, "column_partitions": 1, "inner_partitions": 5, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 1470, "replicas": 294, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 294, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 294, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 294, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 490, "column_partitions": 1, "inner_partitions": 3, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 1470, "replicas": 490, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 490, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 490, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 490, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 490, "column_partitions": 1, "inner_partitions": 3, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 1470, "replicas": 490, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 490, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 490, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 490, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 490, "column_partitions": 1, "inner_partitions": 3, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 1470, "replicas": 490, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 490, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 490, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 490, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 490, "column_partitions": 1, "inner_partitions": 3, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 1470, "replicas": 490, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 490, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 490, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 490, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 245, "column_partitions": 1, "inner_partitions": 6, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 1470, "replicas": 245, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 245, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 245, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 245, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 245, "column_partitions": 1, "inner_partitions": 6, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 1470, "replicas": 245, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 245, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 245, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 245, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 245, "column_partitions": 1, "inner_partitions": 6, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 1470, "replicas": 245, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 245, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 245, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 245, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 245, "column_partitions": 1, "inner_partitions": 6, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 1470, "replicas": 245, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 245, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 245, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 245, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 245, "column_partitions": 1, "inner_partitions": 6, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 1470, "replicas": 245, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 245, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 245, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 245, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 245, "column_partitions": 1, "inner_partitions": 6, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 1470, "replicas": 245, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 245, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 245, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 245, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 245, "column_partitions": 1, "inner_partitions": 6, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 1470, "replicas": 245, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 245, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 245, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 245, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 245, "column_partitions": 1, "inner_partitions": 6, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 1470, "replicas": 245, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 245, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 245, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 245, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 864, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 864, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 864, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 864, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 864, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 864, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 864, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 864, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 184, "column_partitions": 1, "inner_partitions": 8, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 1472, "replicas": 184, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 184, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 184, "column_partitions": 1, "inner_partitions": 8, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 1472, "replicas": 184, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 184, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 184, "column_partitions": 1, "inner_partitions": 8, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 1472, "replicas": 184, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 184, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 184, "column_partitions": 1, "inner_partitions": 8, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 1472, "replicas": 184, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 184, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 184, "column_partitions": 1, "inner_partitions": 8, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 1472, "replicas": 184, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 184, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 184, "column_partitions": 1, "inner_partitions": 8, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 1472, "replicas": 184, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 184, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 184, "column_partitions": 1, "inner_partitions": 8, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 1472, "replicas": 184, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 184, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 184, "column_partitions": 1, "inner_partitions": 8, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 1472, "replicas": 184, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 184, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 163, "column_partitions": 1, "inner_partitions": 9, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 1467, "replicas": 163, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1467, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 163, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 163, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 163, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 163, "column_partitions": 1, "inner_partitions": 9, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 1467, "replicas": 163, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1467, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 163, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 163, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 163, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 163, "column_partitions": 1, "inner_partitions": 9, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 1467, "replicas": 163, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1467, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 163, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 163, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 163, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 163, "column_partitions": 1, "inner_partitions": 9, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 1467, "replicas": 163, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1467, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 163, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 163, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 163, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 81, "column_partitions": 1, "inner_partitions": 18, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 1458, "replicas": 81, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 81, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 81, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 81, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 81, "column_partitions": 1, "inner_partitions": 18, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 1458, "replicas": 81, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 81, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 81, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 81, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 81, "column_partitions": 1, "inner_partitions": 18, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 1458, "replicas": 81, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 81, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 81, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 81, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 81, "column_partitions": 1, "inner_partitions": 18, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 1458, "replicas": 81, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 81, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 81, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 81, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 163, "column_partitions": 1, "inner_partitions": 9, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 1467, "replicas": 163, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1467, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 163, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 163, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 163, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 163, "column_partitions": 1, "inner_partitions": 9, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 1467, "replicas": 163, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1467, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 163, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 163, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 163, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 163, "column_partitions": 1, "inner_partitions": 9, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 1467, "replicas": 163, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1467, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 163, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 163, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 163, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 163, "column_partitions": 1, "inner_partitions": 9, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 1467, "replicas": 163, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1467, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 163, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 163, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 163, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 81, "column_partitions": 1, "inner_partitions": 18, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 1458, "replicas": 81, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 81, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 81, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 81, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 81, "column_partitions": 1, "inner_partitions": 18, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 1458, "replicas": 81, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 81, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 81, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 81, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 81, "column_partitions": 1, "inner_partitions": 18, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 1458, "replicas": 81, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 81, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 81, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 81, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 81, "column_partitions": 1, "inner_partitions": 18, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 1458, "replicas": 81, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 81, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 81, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 81, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 40, "column_partitions": 1, "inner_partitions": 36, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 1440, "replicas": 40, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 40, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 40, "column_partitions": 1, "inner_partitions": 36, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 1440, "replicas": 40, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 40, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 40, "column_partitions": 1, "inner_partitions": 36, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 1440, "replicas": 40, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 40, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 16, "orientation": "Swapped", "distribution": {"ParallelReduction": {"row_partitions": 40, "column_partitions": 1, "inner_partitions": 36, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 1440, "replicas": 40, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 40, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 1}]]}, "deferred_output": null}], "29": [{"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 576, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 288}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 288}, {"axis": {"FromEnd": 1}, "partitions": 288, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 576, "replicas": 2, "axes": [{"axis": {"FromStart": 2}, "partitions": 288, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 576, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 288}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 288}, {"axis": {"FromEnd": 1}, "partitions": 288, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1472, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1472, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1472, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1024, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1024, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1024, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 256}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 256}, {"axis": {"FromEnd": 1}, "partitions": 256, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 2, "axes": [{"axis": {"FromStart": 2}, "partitions": 256, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 256}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 256}, {"axis": {"FromEnd": 1}, "partitions": 256, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 128}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 128}, {"axis": {"FromEnd": 1}, "partitions": 128, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 2, "axes": [{"axis": {"FromStart": 2}, "partitions": 128, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 128}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 128}, {"axis": {"FromEnd": 1}, "partitions": 128, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 64}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 64}, {"axis": {"FromEnd": 1}, "partitions": 64, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 2, "axes": [{"axis": {"FromStart": 2}, "partitions": 64, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 64}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 64}, {"axis": {"FromEnd": 1}, "partitions": 64, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 32, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 2, "axes": [{"axis": {"FromStart": 2}, "partitions": 32, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 32, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 16}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 16}, {"axis": {"FromEnd": 1}, "partitions": 16, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 2, "axes": [{"axis": {"FromStart": 2}, "partitions": 16, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 16}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 16}, {"axis": {"FromEnd": 1}, "partitions": 16, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 8}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 8}, {"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 2, "axes": [{"axis": {"FromStart": 2}, "partitions": 8, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 8}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 8}, {"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 4}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 4}, {"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 2, "axes": [{"axis": {"FromStart": 2}, "partitions": 4, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 4}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 4}, {"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 2}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 2, "axes": [{"axis": {"FromStart": 2}, "partitions": 2, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 2}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 2, "replicas": 2, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1, "replicas": 1, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}], "30": [{"operator": {"Slice": {"axis": 2, "start": 0, "length": 1152}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}], "31": [{"operator": {"Slice": {"axis": 2, "start": 1152, "length": 1152}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}], "32": [{"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 16, "reversed": false}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": {"source_input": 0, "transform": {"split_axis": 2, "merge_axis": 0, "factor": 16, "reversed": false}, "unfused_cycles": 0, "unfused_exchange_cycles": 0}}], "33": [{"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 16, "reversed": false}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": {"source_input": 0, "transform": {"split_axis": 2, "merge_axis": 0, "factor": 16, "reversed": false}, "unfused_cycles": 0, "unfused_exchange_cycles": 0}}], "34": [{"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 16, "reversed": false}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": {"source_input": 0, "transform": {"split_axis": 2, "merge_axis": 0, "factor": 16, "reversed": false}, "unfused_cycles": 0, "unfused_exchange_cycles": 0}}], "35": [{"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": null, "probability_value": null, "materialized": false, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": null, "probability_value": null, "materialized": false, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": null, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 1, "columns": 46, "inner": 1}, "probability_value": null, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": null, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": null, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": null, "probability_value": {"rows": 1, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": null, "probability_value": {"rows": 1, "columns": 5, "inner": 2}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": null, "probability_value": {"rows": 1, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": null, "probability_value": {"rows": 1, "columns": 5, "inner": 3}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": null, "probability_value": {"rows": 1, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": null, "probability_value": {"rows": 1, "columns": 5, "inner": 4}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": null, "probability_value": {"rows": 1, "columns": 5, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": null, "probability_value": {"rows": 1, "columns": 5, "inner": 5}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": null, "probability_value": {"rows": 1, "columns": 5, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": null, "probability_value": {"rows": 1, "columns": 5, "inner": 6}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": null, "probability_value": {"rows": 1, "columns": 5, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": null, "probability_value": {"rows": 1, "columns": 5, "inner": 7}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": null, "probability_value": {"rows": 1, "columns": 5, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": null, "probability_value": {"rows": 1, "columns": 5, "inner": 8}, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": null, "probability_value": null, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": null, "probability_value": null, "materialized": true, "key_block_rows": 768, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 384, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}], "36": [{"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 16, "reversed": true}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}], "37": [{"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 18, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 18, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 18, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 18, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 18, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 18, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 18, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 18, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1296, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 40, "inner_partitions": 36, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 40, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 40, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 40}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 40, "inner_partitions": 36, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 40, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 40, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 40}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 40, "inner_partitions": 36, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 40, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 40, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 40}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 40, "inner_partitions": 36, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 40, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 40, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 40}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 40, "inner_partitions": 36, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 40, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 40, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 40}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 40, "inner_partitions": 36, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 40, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 40, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 40}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 40, "inner_partitions": 36, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 40, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 40, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 40}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 40, "inner_partitions": 36, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 40, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 40, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 40}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 864, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 864, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 864, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 864, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 864, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 864, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 864, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 864, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 864, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 864, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 864, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 864, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 864, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 864, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 864, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 864, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}], "38": [{"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 576, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 288}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 288}, {"axis": {"FromEnd": 1}, "partitions": 288, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 576, "replicas": 2, "axes": [{"axis": {"FromStart": 2}, "partitions": 288, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 576, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 288}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 288}, {"axis": {"FromEnd": 1}, "partitions": 288, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1472, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1472, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1472, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1024, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1024, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1024, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 256}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 256}, {"axis": {"FromEnd": 1}, "partitions": 256, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 2, "axes": [{"axis": {"FromStart": 2}, "partitions": 256, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 256}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 256}, {"axis": {"FromEnd": 1}, "partitions": 256, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 128}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 128}, {"axis": {"FromEnd": 1}, "partitions": 128, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 2, "axes": [{"axis": {"FromStart": 2}, "partitions": 128, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 128}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 128}, {"axis": {"FromEnd": 1}, "partitions": 128, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 64}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 64}, {"axis": {"FromEnd": 1}, "partitions": 64, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 2, "axes": [{"axis": {"FromStart": 2}, "partitions": 64, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 64}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 64}, {"axis": {"FromEnd": 1}, "partitions": 64, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 32, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 2, "axes": [{"axis": {"FromStart": 2}, "partitions": 32, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 32, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 16}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 16}, {"axis": {"FromEnd": 1}, "partitions": 16, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 2, "axes": [{"axis": {"FromStart": 2}, "partitions": 16, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 16}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 16}, {"axis": {"FromEnd": 1}, "partitions": 16, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 8}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 8}, {"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 2, "axes": [{"axis": {"FromStart": 2}, "partitions": 8, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 8}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 8}, {"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 4}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 4}, {"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 2, "axes": [{"axis": {"FromStart": 2}, "partitions": 4, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 4}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 4}, {"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 2}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 2, "axes": [{"axis": {"FromStart": 2}, "partitions": 2, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 2}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 2, "replicas": 2, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1, "replicas": 1, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}], "39": [{"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 2, "replicas": 2, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 2, "replicas": 2, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"LayerNorm": {"parts": 2}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 2}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 2, "axes": [{"axis": {"FromStart": 2}, "partitions": 2, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 2, "axes": [{"axis": {"FromStart": 2}, "partitions": 2, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 2}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"LayerNorm": {"parts": 4}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 4}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 4}, {"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 2, "axes": [{"axis": {"FromStart": 2}, "partitions": 4, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 2, "axes": [{"axis": {"FromStart": 2}, "partitions": 4, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 4}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 4}, {"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"LayerNorm": {"parts": 8}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 8}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 8}, {"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 2, "axes": [{"axis": {"FromStart": 2}, "partitions": 8, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 2, "axes": [{"axis": {"FromStart": 2}, "partitions": 8, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 8}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 8}, {"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"LayerNorm": {"parts": 16}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 16}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 16}, {"axis": {"FromEnd": 1}, "partitions": 16, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 2, "axes": [{"axis": {"FromStart": 2}, "partitions": 16, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 2, "axes": [{"axis": {"FromStart": 2}, "partitions": 16, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 16}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 16}, {"axis": {"FromEnd": 1}, "partitions": 16, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"LayerNorm": {"parts": 32}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 32, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 2, "axes": [{"axis": {"FromStart": 2}, "partitions": 32, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 2, "axes": [{"axis": {"FromStart": 2}, "partitions": 32, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 32, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1472, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1472, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1472, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1472, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1024, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1024, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1024, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1024, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1, "replicas": 1, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1, "replicas": 1, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}, {"operator": "LayerNorm", "dispatch": {"Pointwise": {"kernel": "LayerNorm"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": []}, "deferred_output": null}], "40": [{"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 112, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 40, "inner_partitions": 36, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 40, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 40, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 40}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 112, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 40, "inner_partitions": 36, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 40, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 40, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 40}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 112, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 40, "inner_partitions": 36, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 40, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 40, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 40}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 112, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 40, "inner_partitions": 36, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 40, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 40, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 40}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 112, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 40, "inner_partitions": 36, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 40, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 40, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 40}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 112, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 40, "inner_partitions": 36, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 40, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 40, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 40}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 112, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 40, "inner_partitions": 36, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 40, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 40, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 40}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 112, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 40, "inner_partitions": 36, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 40, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 40, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 40}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 269, "inner_partitions": 5, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1345, "replicas": 269, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 1345, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 269, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 269, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 269, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 269}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 269, "inner_partitions": 5, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1345, "replicas": 269, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 1345, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 269, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 269, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 269, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 269}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 269, "inner_partitions": 5, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1345, "replicas": 269, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 1345, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 269, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 269, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 269, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 269}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 269, "inner_partitions": 5, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1345, "replicas": 269, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 1345, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 269, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 269, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 269, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 269}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 269, "inner_partitions": 5, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1345, "replicas": 269, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 1345, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 269, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 269, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 269, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 269}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 269, "inner_partitions": 5, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1345, "replicas": 269, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 1345, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 269, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 269, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 269, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 269}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 269, "inner_partitions": 5, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1345, "replicas": 269, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 1345, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 269, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 269, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 269, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 269}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 269, "inner_partitions": 5, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1345, "replicas": 269, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 1345, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 269, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 269, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 269, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 269}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 163, "inner_partitions": 9, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1467, "replicas": 163, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 1467, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 163, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 163, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 163, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 163}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 163, "inner_partitions": 9, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1467, "replicas": 163, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 1467, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 163, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 163, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 163, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 163}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 163, "inner_partitions": 9, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1467, "replicas": 163, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 1467, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 163, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 163, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 163, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 163}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 163, "inner_partitions": 9, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1467, "replicas": 163, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 1467, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 163, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 163, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 163, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 163}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 163, "inner_partitions": 9, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1467, "replicas": 163, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 1467, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 163, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 163, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 163, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 163}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 163, "inner_partitions": 9, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1467, "replicas": 163, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 1467, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 163, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 163, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 163, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 163}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 163, "inner_partitions": 9, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1467, "replicas": 163, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 1467, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 163, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 163, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 163, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 163}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 163, "inner_partitions": 9, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1467, "replicas": 163, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 1467, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 163, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 163, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 163, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 163}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 81, "inner_partitions": 18, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 81, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 81, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 81, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 81, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 81}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 81, "inner_partitions": 18, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 81, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 81, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 81, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 81, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 81}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 81, "inner_partitions": 18, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 81, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 81, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 81, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 81, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 81}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 81, "inner_partitions": 18, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 81, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 81, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 81, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 81, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 81}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 81, "inner_partitions": 18, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 81, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 81, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 81, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 81, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 81}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 81, "inner_partitions": 18, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 81, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 81, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 81, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 81, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 81}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 81, "inner_partitions": 18, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 81, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 81, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 81, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 81, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 81}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 81, "inner_partitions": 18, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 81, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 81, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 81, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 81, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 81}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 269, "inner_partitions": 4, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1076, "replicas": 269, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 1076, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 269, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 269, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 269, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 269}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 269, "inner_partitions": 4, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1076, "replicas": 269, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 1076, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 269, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 269, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 269, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 269}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 269, "inner_partitions": 4, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1076, "replicas": 269, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 1076, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 269, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 269, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 269, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 269}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 269, "inner_partitions": 4, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1076, "replicas": 269, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 1076, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 269, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 269, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 269, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 269}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 269, "inner_partitions": 4, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1076, "replicas": 269, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 1076, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 269, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 269, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 269, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 269}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 269, "inner_partitions": 4, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1076, "replicas": 269, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 1076, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 269, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 269, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 269, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 269}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 269, "inner_partitions": 4, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1076, "replicas": 269, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 1076, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 269, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 269, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 269, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 269}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 269, "inner_partitions": 4, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1076, "replicas": 269, "axes": [{"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 1076, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 269, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 269, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 269, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 269}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 122, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1464, "replicas": 122, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 1464, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 122, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 122}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 122, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1464, "replicas": 122, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 1464, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 122, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 122}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 122, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1464, "replicas": 122, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 1464, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 122, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 122}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 122, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1464, "replicas": 122, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 1464, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 122, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 122}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 122, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1464, "replicas": 122, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 1464, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 122, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 122}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 122, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1464, "replicas": 122, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 1464, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 122, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 122}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 122, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1464, "replicas": 122, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 1464, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 122, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 122}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 122, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1464, "replicas": 122, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 1464, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 122, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 122, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 122}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 184, "inner_partitions": 8, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 184, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 184, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 184}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 184, "inner_partitions": 8, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 184, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 184, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 184}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 184, "inner_partitions": 8, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 184, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 184, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 184}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 184, "inner_partitions": 8, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 184, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 184, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 184}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 184, "inner_partitions": 8, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 184, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 184, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 184}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 184, "inner_partitions": 8, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 184, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 184, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 184}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 184, "inner_partitions": 8, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 184, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 184, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 184}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 184, "inner_partitions": 8, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 184, "axes": [{"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 184, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 184, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 184}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 269, "inner_partitions": 3, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 807, "replicas": 269, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 807, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 269, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 269, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 269, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 269}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 269, "inner_partitions": 3, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 807, "replicas": 269, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 807, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 269, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 269, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 269, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 269}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 269, "inner_partitions": 3, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 807, "replicas": 269, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 807, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 269, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 269, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 269, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 269}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 269, "inner_partitions": 3, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 807, "replicas": 269, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 807, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 269, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 269, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 269, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 269}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 269, "inner_partitions": 3, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 807, "replicas": 269, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 807, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 269, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 269, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 269, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 269}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 269, "inner_partitions": 3, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 807, "replicas": 269, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 807, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 269, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 269, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 269, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 269}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 269, "inner_partitions": 3, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 807, "replicas": 269, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 807, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 269, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 269, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 269, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 269}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 269, "inner_partitions": 3, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 807, "replicas": 269, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 807, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 269, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 269, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 269, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 269}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 245, "inner_partitions": 6, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 245, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 245, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 245, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 245, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 245}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 245, "inner_partitions": 6, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 245, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 245, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 245, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 245, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 245}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 245, "inner_partitions": 6, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 245, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 245, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 245, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 245, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 245}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 245, "inner_partitions": 6, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 245, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 245, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 245, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 245, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 245}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 245, "inner_partitions": 6, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 245, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 245, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 245, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 245, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 245}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 245, "inner_partitions": 6, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 245, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 245, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 245, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 245, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 245}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 245, "inner_partitions": 6, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 245, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 245, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 245, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 245, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 245}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 245, "inner_partitions": 6, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1470, "replicas": 245, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 1470, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 245, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 245, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 245, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 245}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 112, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 40, "inner_partitions": 36, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 40, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 1404, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 39, "padding_groups": 1, "block_size": 112, "padding_multiple": 112, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 40, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 40}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 112, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 40, "inner_partitions": 36, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 40, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 1404, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 39, "padding_groups": 1, "block_size": 112, "padding_multiple": 112, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 40, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 40}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 112, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 40, "inner_partitions": 36, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 40, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 1404, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 39, "padding_groups": 1, "block_size": 112, "padding_multiple": 112, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 40, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 40}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 112, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 40, "inner_partitions": 36, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 40, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 1404, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 39, "padding_groups": 1, "block_size": 112, "padding_multiple": 112, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 40, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 40}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 112, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 40, "inner_partitions": 36, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 40, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 1404, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 39, "padding_groups": 1, "block_size": 112, "padding_multiple": 112, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 40, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 40}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 112, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 40, "inner_partitions": 36, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 40, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 1404, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 39, "padding_groups": 1, "block_size": 112, "padding_multiple": 112, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 40, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 40}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 112, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 40, "inner_partitions": 36, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 40, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 1404, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 39, "padding_groups": 1, "block_size": 112, "padding_multiple": 112, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 40, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 40}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 112, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 40, "inner_partitions": 36, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 40, "axes": [{"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 1404, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 39, "padding_groups": 1, "block_size": 112, "padding_multiple": 112, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 40, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 40}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 81, "inner_partitions": 18, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 81, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 1224, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 68, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 81, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 81, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 81}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 81, "inner_partitions": 18, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 81, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 1224, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 68, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 81, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 81, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 81}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 81, "inner_partitions": 18, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 81, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 1224, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 68, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 81, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 81, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 81}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 81, "inner_partitions": 18, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 81, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 1224, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 68, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 81, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 81, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 81}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 81, "inner_partitions": 18, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 81, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 1224, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 68, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 81, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 81, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 81}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 81, "inner_partitions": 18, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 81, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 1224, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 68, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 81, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 81, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 81}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 81, "inner_partitions": 18, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 81, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 1224, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 68, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 81, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 81, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 81}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 81, "inner_partitions": 18, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 81, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 1224, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 68, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 81, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 81, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 81}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}], "41": [{"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 736}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 736}, {"axis": {"FromEnd": 1}, "partitions": 736, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 2, "axes": [{"axis": {"FromStart": 2}, "partitions": 736, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 736}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 736}, {"axis": {"FromEnd": 1}, "partitions": 736, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1472, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1472, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1472, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 512}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 512}, {"axis": {"FromEnd": 1}, "partitions": 512, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 2, "axes": [{"axis": {"FromStart": 2}, "partitions": 512, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 512}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 512}, {"axis": {"FromEnd": 1}, "partitions": 512, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1024, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1024, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1024, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 256}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 256}, {"axis": {"FromEnd": 1}, "partitions": 256, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 2, "axes": [{"axis": {"FromStart": 2}, "partitions": 256, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 256}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 256}, {"axis": {"FromEnd": 1}, "partitions": 256, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 128}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 128}, {"axis": {"FromEnd": 1}, "partitions": 128, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 2, "axes": [{"axis": {"FromStart": 2}, "partitions": 128, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 128}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 128}, {"axis": {"FromEnd": 1}, "partitions": 128, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 64}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 64}, {"axis": {"FromEnd": 1}, "partitions": 64, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 2, "axes": [{"axis": {"FromStart": 2}, "partitions": 64, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 64}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 64}, {"axis": {"FromEnd": 1}, "partitions": 64, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 32, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 2, "axes": [{"axis": {"FromStart": 2}, "partitions": 32, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 32, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 16}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 16}, {"axis": {"FromEnd": 1}, "partitions": 16, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 2, "axes": [{"axis": {"FromStart": 2}, "partitions": 16, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 16}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 16}, {"axis": {"FromEnd": 1}, "partitions": 16, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 8}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 8}, {"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 2, "axes": [{"axis": {"FromStart": 2}, "partitions": 8, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 8}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 8}, {"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 4}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 4}, {"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 2, "axes": [{"axis": {"FromStart": 2}, "partitions": 4, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 4}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 4}, {"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 2}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 2, "axes": [{"axis": {"FromStart": 2}, "partitions": 2, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 2}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 2, "replicas": 2, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1, "replicas": 1, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}], "42": [{"operator": "Gelu", "dispatch": {"Pointwise": {"kernel": "Gelu"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 2, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 2, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Gelu", "dispatch": {"Pointwise": {"kernel": "Gelu"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 4, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 4, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Gelu", "dispatch": {"Pointwise": {"kernel": "Gelu"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 8, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 8, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Gelu", "dispatch": {"Pointwise": {"kernel": "Gelu"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 16, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 16, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Gelu", "dispatch": {"Pointwise": {"kernel": "Gelu"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 32, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 32, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Gelu", "dispatch": {"Pointwise": {"kernel": "Gelu"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 64, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 64, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Gelu", "dispatch": {"Pointwise": {"kernel": "Gelu"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 128, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 128, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Gelu", "dispatch": {"Pointwise": {"kernel": "Gelu"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 256, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 256, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Gelu", "dispatch": {"Pointwise": {"kernel": "Gelu"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 512, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 512, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Gelu", "dispatch": {"Pointwise": {"kernel": "Gelu"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 1024, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 1024, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Gelu", "dispatch": {"Pointwise": {"kernel": "Gelu"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 1472, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 1472, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Gelu", "dispatch": {"Pointwise": {"kernel": "Gelu"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}], "43": [{"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 224, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 20, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 20, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 224, "column_block": 16}}}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 224, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 20, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 20, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 224, "column_block": 16}}}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 224, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 20, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 20, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 224, "column_block": 16}}}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 224, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 20, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 20, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 224, "column_block": 16}}}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 224, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 20, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 20, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 224, "column_block": 16}}}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 224, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 20, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 20, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 224, "column_block": 16}}}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 224, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 20, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 20, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 224, "column_block": 16}}}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 224, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 20, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 20, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 224, "column_block": 16}}}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 17, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1224, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 17, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 1224, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 17, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 17, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1224, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 17, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 1224, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 17, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 17, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1224, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 17, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 1224, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 17, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 17, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1224, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 17, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 1224, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 17, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 17, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1224, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 17, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 1224, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 17, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 17, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1224, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 17, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 1224, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 17, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 17, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1224, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 17, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 1224, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 17, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 256, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 17, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1224, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 17, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 256, "column_block": 16}}}, "tiling": {"tile_count": 1224, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 17, "padding_groups": 1, "block_size": 256, "padding_multiple": 256, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 43, "inner_partitions": 34, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1462, "replicas": 43, "axes": [{"axis": {"FromEnd": 1}, "partitions": 34, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 1462, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 43, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 34, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 43, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 43, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 43}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 43, "inner_partitions": 34, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1462, "replicas": 43, "axes": [{"axis": {"FromEnd": 1}, "partitions": 34, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 1462, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 43, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 34, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 43, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 43, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 43}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 43, "inner_partitions": 34, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1462, "replicas": 43, "axes": [{"axis": {"FromEnd": 1}, "partitions": 34, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 1462, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 43, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 34, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 43, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 43, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 43}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 43, "inner_partitions": 34, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1462, "replicas": 43, "axes": [{"axis": {"FromEnd": 1}, "partitions": 34, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 1462, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 43, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 34, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 43, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 43, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 43}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 43, "inner_partitions": 34, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1462, "replicas": 43, "axes": [{"axis": {"FromEnd": 1}, "partitions": 34, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 1462, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 43, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 34, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 43, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 43, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 43}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 43, "inner_partitions": 34, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1462, "replicas": 43, "axes": [{"axis": {"FromEnd": 1}, "partitions": 34, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 1462, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 43, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 34, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 43, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 43, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 43}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 43, "inner_partitions": 34, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1462, "replicas": 43, "axes": [{"axis": {"FromEnd": 1}, "partitions": 34, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 1462, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 43, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 34, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 43, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 43, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 43}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 128, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 43, "inner_partitions": 34, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1462, "replicas": 43, "axes": [{"axis": {"FromEnd": 1}, "partitions": 34, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 1462, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 43, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 34, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 43, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 43, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 43}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 21, "inner_partitions": 68, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1428, "replicas": 21, "axes": [{"axis": {"FromEnd": 1}, "partitions": 68, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 1428, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 21, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 68, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 21, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 21, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 21}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 21, "inner_partitions": 68, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1428, "replicas": 21, "axes": [{"axis": {"FromEnd": 1}, "partitions": 68, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 1428, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 21, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 68, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 21, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 21, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 21}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 21, "inner_partitions": 68, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1428, "replicas": 21, "axes": [{"axis": {"FromEnd": 1}, "partitions": 68, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 1428, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 21, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 68, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 21, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 21, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 21}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 21, "inner_partitions": 68, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1428, "replicas": 21, "axes": [{"axis": {"FromEnd": 1}, "partitions": 68, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 1428, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 21, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 68, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 21, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 21, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 21}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 21, "inner_partitions": 68, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1428, "replicas": 21, "axes": [{"axis": {"FromEnd": 1}, "partitions": 68, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 1428, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 21, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 68, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 21, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 21, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 21}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 21, "inner_partitions": 68, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1428, "replicas": 21, "axes": [{"axis": {"FromEnd": 1}, "partitions": 68, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 1428, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 21, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 68, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 21, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 21, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 21}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 21, "inner_partitions": 68, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1428, "replicas": 21, "axes": [{"axis": {"FromEnd": 1}, "partitions": 68, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 1428, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 21, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 68, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 21, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 21, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 21}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 21, "inner_partitions": 68, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1428, "replicas": 21, "axes": [{"axis": {"FromEnd": 1}, "partitions": 68, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 1428, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 21, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 68, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 21, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 21, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 21}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 128, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 10, "inner_partitions": 135, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1350, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 135, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 1350, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 10, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 135, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 10, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 10, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 10}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 128, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 10, "inner_partitions": 135, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1350, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 135, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 1350, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 10, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 135, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 10, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 10, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 10}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 128, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 10, "inner_partitions": 135, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1350, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 135, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 1350, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 10, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 135, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 10, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 10, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 10}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 128, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 10, "inner_partitions": 135, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1350, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 135, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 1350, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 10, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 135, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 10, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 10, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 10}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 128, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 10, "inner_partitions": 135, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1350, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 135, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 1350, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 10, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 135, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 10, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 10, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 10}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 128, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 10, "inner_partitions": 135, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1350, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 135, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 1350, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 10, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 135, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 10, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 10, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 10}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 128, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 10, "inner_partitions": 135, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1350, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 135, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 1350, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 10, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 135, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 10, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 10, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 10}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 128, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 10, "inner_partitions": 135, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1350, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 135, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 1350, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 10, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 135, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 10, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 10, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 10}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 15, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1080, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 15, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 1080, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 15, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 15, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1080, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 15, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 1080, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 15, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 15, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1080, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 15, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 1080, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 15, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 15, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1080, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 15, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 1080, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 15, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 15, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1080, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 15, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 1080, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 15, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 15, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1080, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 15, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 1080, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 15, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 15, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1080, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 15, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 1080, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 15, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 288, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 15, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1080, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 15, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 288, "column_block": 16}}}, "tiling": {"tile_count": 1080, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 15, "padding_groups": 1, "block_size": 288, "padding_multiple": 288, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 32, "inner_partitions": 45, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 32, "axes": [{"axis": {"FromEnd": 1}, "partitions": 45, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 32, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 45, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 32, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 32, "inner_partitions": 45, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 32, "axes": [{"axis": {"FromEnd": 1}, "partitions": 45, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 32, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 45, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 32, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 32, "inner_partitions": 45, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 32, "axes": [{"axis": {"FromEnd": 1}, "partitions": 45, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 32, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 45, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 32, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 32, "inner_partitions": 45, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 32, "axes": [{"axis": {"FromEnd": 1}, "partitions": 45, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 32, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 45, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 32, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 32, "inner_partitions": 45, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 32, "axes": [{"axis": {"FromEnd": 1}, "partitions": 45, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 32, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 45, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 32, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 32, "inner_partitions": 45, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 32, "axes": [{"axis": {"FromEnd": 1}, "partitions": 45, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 32, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 45, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 32, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 32, "inner_partitions": 45, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 32, "axes": [{"axis": {"FromEnd": 1}, "partitions": 45, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 32, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 45, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 32, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 96, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 32, "inner_partitions": 45, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1440, "replicas": 32, "axes": [{"axis": {"FromEnd": 1}, "partitions": 45, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 32, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 45, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 32, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 32}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 14, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1008, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 14, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 1008, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 14, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 14, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1008, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 14, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 1008, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 14, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 14, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1008, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 14, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 1008, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 14, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 14, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1008, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 14, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 1008, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 14, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 14, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1008, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 14, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 1008, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 14, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 14, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1008, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 14, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 1008, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 14, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 14, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1008, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 14, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 1008, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 14, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 320, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 14, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1008, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 14, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 320, "column_block": 16}}}, "tiling": {"tile_count": 1008, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 14, "padding_groups": 1, "block_size": 320, "padding_multiple": 320, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 54, "inner_partitions": 27, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 54, "axes": [{"axis": {"FromEnd": 1}, "partitions": 27, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 27, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 54, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 54}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 54, "inner_partitions": 27, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 54, "axes": [{"axis": {"FromEnd": 1}, "partitions": 27, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 27, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 54, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 54}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 54, "inner_partitions": 27, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 54, "axes": [{"axis": {"FromEnd": 1}, "partitions": 27, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 27, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 54, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 54}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 54, "inner_partitions": 27, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 54, "axes": [{"axis": {"FromEnd": 1}, "partitions": 27, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 27, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 54, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 54}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 54, "inner_partitions": 27, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 54, "axes": [{"axis": {"FromEnd": 1}, "partitions": 27, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 27, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 54, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 54}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 54, "inner_partitions": 27, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 54, "axes": [{"axis": {"FromEnd": 1}, "partitions": 27, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 27, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 54, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 54}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 54, "inner_partitions": 27, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 54, "axes": [{"axis": {"FromEnd": 1}, "partitions": 27, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 27, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 54, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 54}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 160, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 54, "inner_partitions": 27, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1458, "replicas": 54, "axes": [{"axis": {"FromEnd": 1}, "partitions": 27, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 1458, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 27, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 54, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 54, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 54}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 352, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 13, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 936, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 13, "padding_groups": 1, "block_size": 352, "padding_multiple": 352, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 352, "column_block": 16}}}, "tiling": {"tile_count": 936, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 13, "padding_groups": 1, "block_size": 352, "padding_multiple": 352, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 352, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 13, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 936, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 13, "padding_groups": 1, "block_size": 352, "padding_multiple": 352, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 352, "column_block": 16}}}, "tiling": {"tile_count": 936, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 13, "padding_groups": 1, "block_size": 352, "padding_multiple": 352, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 352, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 13, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 936, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 13, "padding_groups": 1, "block_size": 352, "padding_multiple": 352, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 352, "column_block": 16}}}, "tiling": {"tile_count": 936, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 13, "padding_groups": 1, "block_size": 352, "padding_multiple": 352, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 352, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 13, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 936, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 13, "padding_groups": 1, "block_size": 352, "padding_multiple": 352, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 352, "column_block": 16}}}, "tiling": {"tile_count": 936, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 13, "padding_groups": 1, "block_size": 352, "padding_multiple": 352, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 352, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 13, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 936, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 13, "padding_groups": 1, "block_size": 352, "padding_multiple": 352, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 352, "column_block": 16}}}, "tiling": {"tile_count": 936, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 13, "padding_groups": 1, "block_size": 352, "padding_multiple": 352, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 352, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 13, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 936, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 13, "padding_groups": 1, "block_size": 352, "padding_multiple": 352, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 352, "column_block": 16}}}, "tiling": {"tile_count": 936, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 13, "padding_groups": 1, "block_size": 352, "padding_multiple": 352, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 352, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 13, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 936, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 13, "padding_groups": 1, "block_size": 352, "padding_multiple": 352, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 352, "column_block": 16}}}, "tiling": {"tile_count": 936, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 13, "padding_groups": 1, "block_size": 352, "padding_multiple": 352, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 352, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 13, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 936, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 13, "padding_groups": 1, "block_size": 352, "padding_multiple": 352, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 352, "column_block": 16}}}, "tiling": {"tile_count": 936, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 13, "padding_groups": 1, "block_size": 352, "padding_multiple": 352, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 864, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 864, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 864, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 864, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 864, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 864, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 864, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 864, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 864, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 864, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 864, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 864, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 864, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 864, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 384, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 12, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 864, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 384, "column_block": 16}}}, "tiling": {"tile_count": 864, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 12, "padding_groups": 1, "block_size": 384, "padding_multiple": 384, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 64, "inner_partitions": 23, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 64, "axes": [{"axis": {"FromEnd": 1}, "partitions": 23, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 64, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 23, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 64, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 64}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 64, "inner_partitions": 23, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 64, "axes": [{"axis": {"FromEnd": 1}, "partitions": 23, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 64, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 23, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 64, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 64}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 64, "inner_partitions": 23, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 64, "axes": [{"axis": {"FromEnd": 1}, "partitions": 23, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 64, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 23, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 64, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 64}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 64, "inner_partitions": 23, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 64, "axes": [{"axis": {"FromEnd": 1}, "partitions": 23, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 64, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 23, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 64, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 64}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 64, "inner_partitions": 23, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 64, "axes": [{"axis": {"FromEnd": 1}, "partitions": 23, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 64, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 23, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 64, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 64}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 64, "inner_partitions": 23, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 64, "axes": [{"axis": {"FromEnd": 1}, "partitions": 23, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 64, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 23, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 64, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 64}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 64, "inner_partitions": 23, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 64, "axes": [{"axis": {"FromEnd": 1}, "partitions": 23, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 64, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 23, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 64, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 64}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 192, "output_column_block": 32, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 64, "inner_partitions": 23, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1472, "replicas": 64, "axes": [{"axis": {"FromEnd": 1}, "partitions": 23, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 192, "column_block": 16}}}, "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 64, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 23, "padding_groups": 1, "block_size": 192, "padding_multiple": 192, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 64, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 64}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 416, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 11, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 792, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 11, "padding_groups": 1, "block_size": 416, "padding_multiple": 416, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 416, "column_block": 16}}}, "tiling": {"tile_count": 792, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 11, "padding_groups": 1, "block_size": 416, "padding_multiple": 416, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 416, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 11, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 792, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 11, "padding_groups": 1, "block_size": 416, "padding_multiple": 416, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 416, "column_block": 16}}}, "tiling": {"tile_count": 792, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 11, "padding_groups": 1, "block_size": 416, "padding_multiple": 416, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 416, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 11, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 792, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 11, "padding_groups": 1, "block_size": 416, "padding_multiple": 416, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 416, "column_block": 16}}}, "tiling": {"tile_count": 792, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 11, "padding_groups": 1, "block_size": 416, "padding_multiple": 416, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 416, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 11, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 792, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 11, "padding_groups": 1, "block_size": 416, "padding_multiple": 416, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 416, "column_block": 16}}}, "tiling": {"tile_count": 792, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 11, "padding_groups": 1, "block_size": 416, "padding_multiple": 416, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 416, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 11, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 792, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 11, "padding_groups": 1, "block_size": 416, "padding_multiple": 416, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 416, "column_block": 16}}}, "tiling": {"tile_count": 792, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 11, "padding_groups": 1, "block_size": 416, "padding_multiple": 416, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 416, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 11, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 792, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 11, "padding_groups": 1, "block_size": 416, "padding_multiple": 416, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 416, "column_block": 16}}}, "tiling": {"tile_count": 792, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 11, "padding_groups": 1, "block_size": 416, "padding_multiple": 416, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 416, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 11, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 792, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 11, "padding_groups": 1, "block_size": 416, "padding_multiple": 416, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 416, "column_block": 16}}}, "tiling": {"tile_count": 792, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 11, "padding_groups": 1, "block_size": 416, "padding_multiple": 416, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 416, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 72, "inner_partitions": 11, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 792, "replicas": 72, "axes": [{"axis": {"FromEnd": 1}, "partitions": 11, "padding_groups": 1, "block_size": 416, "padding_multiple": 416, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 416, "column_block": 16}}}, "tiling": {"tile_count": 792, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 11, "padding_groups": 1, "block_size": 416, "padding_multiple": 416, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 72}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 21, "inner_partitions": 68, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1428, "replicas": 21, "axes": [{"axis": {"FromEnd": 1}, "partitions": 68, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 1224, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 68, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 21, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 21, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 21}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 21, "inner_partitions": 68, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1428, "replicas": 21, "axes": [{"axis": {"FromEnd": 1}, "partitions": 68, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 1224, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 68, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 21, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 21, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 21}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 21, "inner_partitions": 68, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1428, "replicas": 21, "axes": [{"axis": {"FromEnd": 1}, "partitions": 68, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 1224, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 68, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 21, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 21, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 21}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 21, "inner_partitions": 68, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1428, "replicas": 21, "axes": [{"axis": {"FromEnd": 1}, "partitions": 68, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 1224, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 68, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 21, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 21, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 21}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 21, "inner_partitions": 68, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1428, "replicas": 21, "axes": [{"axis": {"FromEnd": 1}, "partitions": 68, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 1224, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 68, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 21, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 21, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 21}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 21, "inner_partitions": 68, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1428, "replicas": 21, "axes": [{"axis": {"FromEnd": 1}, "partitions": 68, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 1224, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 68, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 21, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 21, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 21}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 21, "inner_partitions": 68, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1428, "replicas": 21, "axes": [{"axis": {"FromEnd": 1}, "partitions": 68, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 1224, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 68, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 21, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 21, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 21}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 64, "output_column_block": 64, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 21, "inner_partitions": 68, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1428, "replicas": 21, "axes": [{"axis": {"FromEnd": 1}, "partitions": 68, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 1224, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 68, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 21, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 21, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 21}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 128, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 10, "inner_partitions": 135, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1350, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 135, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 1215, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 135, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 10, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 10, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 10}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 128, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 10, "inner_partitions": 135, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1350, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 135, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 1215, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 135, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 10, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 10, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 10}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 128, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 10, "inner_partitions": 135, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1350, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 135, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 1215, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 135, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 10, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 10, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 10}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 128, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 10, "inner_partitions": 135, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1350, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 135, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 1215, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 135, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 10, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 10, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 10}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 128, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 10, "inner_partitions": 135, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1350, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 135, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 1215, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 135, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 10, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 10, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 10}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 128, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 10, "inner_partitions": 135, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Complete"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1350, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 135, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 1215, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 135, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 10, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 10, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 10}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 128, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 10, "inner_partitions": 135, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1350, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 135, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 1215, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 135, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 10, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 10, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 10}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 128, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 10, "inner_partitions": 135, "result_row_partitions": 1, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 1350, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 135, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 32, "access_tail_bytes": 16, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 1215, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 135, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 10, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 10, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 10}]}, "memory_class": "Ipu21Interleaved"}}, "alignment": 32, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh", "distinct_elements": [["Output", {"Input": 0}]]}, "deferred_output": null}], "44": [{"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 576, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 288}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 288}, {"axis": {"FromEnd": 1}, "partitions": 288, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 576, "replicas": 2, "axes": [{"axis": {"FromStart": 2}, "partitions": 288, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 576, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 288}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 288}, {"axis": {"FromEnd": 1}, "partitions": 288, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1472, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1472, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1472, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1024, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1024, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1024, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 256}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 256}, {"axis": {"FromEnd": 1}, "partitions": 256, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 2, "axes": [{"axis": {"FromStart": 2}, "partitions": 256, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 256}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 256}, {"axis": {"FromEnd": 1}, "partitions": 256, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 128}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 128}, {"axis": {"FromEnd": 1}, "partitions": 128, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 2, "axes": [{"axis": {"FromStart": 2}, "partitions": 128, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 128}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 128}, {"axis": {"FromEnd": 1}, "partitions": 128, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 64}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 64}, {"axis": {"FromEnd": 1}, "partitions": 64, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 2, "axes": [{"axis": {"FromStart": 2}, "partitions": 64, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 64}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 64}, {"axis": {"FromEnd": 1}, "partitions": 64, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 32, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 2, "axes": [{"axis": {"FromStart": 2}, "partitions": 32, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 32, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 16}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 16}, {"axis": {"FromEnd": 1}, "partitions": 16, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 2, "axes": [{"axis": {"FromStart": 2}, "partitions": 16, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 16}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 16}, {"axis": {"FromEnd": 1}, "partitions": 16, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 8}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 8}, {"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 2, "axes": [{"axis": {"FromStart": 2}, "partitions": 8, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 8}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 8}, {"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 4}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 4}, {"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 2, "axes": [{"axis": {"FromStart": 2}, "partitions": 4, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 4}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 4}, {"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 2}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 2, "axes": [{"axis": {"FromStart": 2}, "partitions": 2, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 2}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 2, "replicas": 2, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1, "replicas": 1, "axes": [{"axis": {"FromStart": 2}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}, "distinct_elements": []}, "deferred_output": null}], "45": [{"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 576, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 288}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 288}, {"axis": {"FromEnd": 1}, "partitions": 288, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 576, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 288}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 288}, {"axis": {"FromEnd": 1}, "partitions": 288, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 576, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 288}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 288}, {"axis": {"FromEnd": 1}, "partitions": 288, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1472, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1472, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1472, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1024, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1024, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1024, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1024, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 256}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 256}, {"axis": {"FromEnd": 1}, "partitions": 256, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 256}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 256}, {"axis": {"FromEnd": 1}, "partitions": 256, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 256}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 256}, {"axis": {"FromEnd": 1}, "partitions": 256, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 512, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 512, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 128}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 128}, {"axis": {"FromEnd": 1}, "partitions": 128, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 128}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 128}, {"axis": {"FromEnd": 1}, "partitions": 128, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 128}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 128}, {"axis": {"FromEnd": 1}, "partitions": 128, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 256, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 256, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 64}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 64}, {"axis": {"FromEnd": 1}, "partitions": 64, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 64}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 64}, {"axis": {"FromEnd": 1}, "partitions": 64, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 64}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 64}, {"axis": {"FromEnd": 1}, "partitions": 64, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 128, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 128, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 32, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 32, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 32}, {"axis": {"FromEnd": 1}, "partitions": 32, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 64, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 16}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 16}, {"axis": {"FromEnd": 1}, "partitions": 16, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 16}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 16}, {"axis": {"FromEnd": 1}, "partitions": 16, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 16}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 16}, {"axis": {"FromEnd": 1}, "partitions": 16, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 32, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 8}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 8}, {"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 8}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 8}, {"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 8}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 8}, {"axis": {"FromEnd": 1}, "partitions": 8, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 16, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 4}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 4}, {"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 4}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 4}, {"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 4}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 4}, {"axis": {"FromEnd": 1}, "partitions": 4, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 8, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 2}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 2}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 2}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 4, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 3}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "alignment": 8, "access_tail_bytes": 0, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}, "distinct_elements": []}, "deferred_output": null}]}, "inputs": {"0": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 729, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 729, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "1": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1368, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 19, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "2": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "5": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1472, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 1472, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "7": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "8": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "9": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "10": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "11": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "12": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "13": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "14": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "15": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "16": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "17": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "18": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "19": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "20": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "21": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "22": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "23": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "24": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "25": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "26": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "27": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "28": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "29": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "30": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "31": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "32": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "33": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "34": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "35": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "36": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "37": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "38": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "39": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "40": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "41": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "42": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "43": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "44": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "45": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "46": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "47": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "48": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "49": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "50": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "51": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "52": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "53": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "54": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "55": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "56": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "57": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "58": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "59": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "60": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "61": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "62": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "63": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "64": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "65": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "66": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "67": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "68": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "69": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "70": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "71": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "72": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "73": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "74": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "75": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "76": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "77": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "78": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "79": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "80": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "81": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "82": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "83": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "84": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "85": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "86": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "87": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 6, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 216, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "88": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 27, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 27, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "89": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 27, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 27, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "90": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 27, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 27, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "91": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 27, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 27, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "92": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 27, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 27, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "93": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 27, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 27, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "94": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 27, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 27, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "95": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 27, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 27, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "96": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 27, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 27, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "97": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 27, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 27, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "98": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 27, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 27, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "99": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 27, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 27, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "100": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 27, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 27, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "101": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 27, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 27, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "102": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 27, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 27, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "103": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 27, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 27, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "104": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 27, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 27, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "105": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 27, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 27, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "106": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 27, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 27, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "107": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 27, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 27, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "108": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 27, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 27, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "109": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 27, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 27, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "110": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 27, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 27, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "111": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 27, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 27, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "112": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 27, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 27, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "113": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 27, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 27, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "114": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 27, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 27, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "115": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "116": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "117": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "118": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "119": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "120": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "121": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "122": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "123": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "124": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "125": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "126": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "127": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "128": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "129": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "130": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "131": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "132": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "133": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "134": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "135": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "136": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "137": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "138": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "139": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "140": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "141": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "142": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "143": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "144": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "145": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "146": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "147": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "148": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "149": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "150": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "151": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "152": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "153": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "154": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "155": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "156": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "157": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "158": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "159": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "160": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "161": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "162": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "163": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "164": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "165": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "166": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "167": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "168": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "169": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "170": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "171": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "172": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "173": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "174": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "175": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "176": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "177": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "178": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "179": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "180": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "181": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "182": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "183": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "184": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "185": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "186": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "187": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "188": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "189": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "190": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "191": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "192": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "193": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "194": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "195": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "196": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "197": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "198": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "199": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "200": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "201": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "202": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "203": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "204": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "205": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "206": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "207": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "208": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "209": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "210": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "211": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "212": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "213": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "214": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "215": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "216": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "217": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "218": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "219": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "220": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "221": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "222": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "223": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "224": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "225": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "226": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "227": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "228": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "229": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "230": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "231": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "232": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "233": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "234": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "235": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "236": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "237": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "238": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "239": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "240": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "241": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "242": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "243": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "244": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "245": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "246": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "247": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "248": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "249": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "250": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 34, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 34, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "251": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 34, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 34, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "252": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 34, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 34, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "253": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 34, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 34, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "254": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 34, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 34, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "255": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 34, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 34, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "256": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 34, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 34, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "257": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 34, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 34, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "258": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 34, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 34, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "259": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 34, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 34, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "260": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 34, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 34, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "261": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 34, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 34, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "262": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 34, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 34, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "263": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 34, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 34, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "264": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 34, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 34, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "265": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 34, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 34, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "266": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 34, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 34, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "267": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 34, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 34, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "268": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 34, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 34, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "269": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 34, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 34, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "270": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 34, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 34, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "271": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 34, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 34, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "272": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 34, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 34, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "273": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 34, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 34, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "274": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 34, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 34, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "275": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 34, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 34, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "276": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 34, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 34, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "277": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "278": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "279": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "280": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "281": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "282": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "283": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "284": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "285": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "286": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "287": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "288": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "289": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "290": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "291": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "292": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "293": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "294": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "295": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "296": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "297": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "298": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "299": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "300": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "301": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "302": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "303": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "304": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "305": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "306": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "307": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "308": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "309": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "310": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "311": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "312": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "313": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "314": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "315": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "316": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "317": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "318": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "319": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "320": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "321": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "322": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "323": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "324": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "325": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "326": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "327": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "328": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "329": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "330": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "366": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "367": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "369": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 72, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "370": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 9, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 144, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "371": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 18, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 18, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "374": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "375": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "385": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1296, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 18, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "386": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "389": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "390": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "392": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 36, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 40, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "393": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 34, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 34, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "397": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 1440, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 20, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 72, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "398": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 9, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 9, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}}, "visited": [], "mapping": null, "mapping_checked": true, "attempts": 0}